Feedback on BTest-1 Unit

[Jim Gettys]

I think this path is the wrong place to concentrate right now.

We should be working on making suspend/resume to/from RAM work, and go
as fast as possible.  If others want to work on suspend to disk, fine,
but boot time isn't interesting once you have fast suspend/resume
working with good power management.
                                  Regards,
                                         - Jim


On Mon, 2006-12-04 at 13:32 -0700, Jordan Crouse wrote:
> On 04/12/06 13:16 -0600, Richard Smith wrote:
> > Chris Ball wrote:
> > 
> > > Mitch points out that jffs2 takes a few seconds to mount, so we might
> > > partition the NAND, and use FAT for a small partition with this memory
> > > dump in.  Our startup time could drop to just a few seconds, and we
> > > could compress the RAM image.
> > > 
> > > Has anyone worked on anything like this before?  Any pointers?
> > 
> > The suspend2 patches do most of this already.  You can specify a default
> > suspend image that gets loaded on boot.  Kernel help says its for use in
> > embedded systems to make the system come back up in a default loaded state.
> 
> Thats the general gist of the OLS talk.  The challenge is in figuring out
> the optimal image.  We don't want to include X and sugar in the image due to
> the keyboard, language and font differences - and I would argue that all of
> userland is probably not useful to us.  So then we we'll need to take care
> to only snapshot a pristine version of the kernel memory without anything
> plugged into USB, etc, etc.  Its not impossible, and its probably practical
> with suspend2, but like you said, its going to take tweaking.
> 
> Definitely something that should be considered though - could we start by 
> querying Nigel Cunningham and the other suspend2 folks for some advice?
> 
> Jordan
> 
-- 
Jim Gettys
One Laptop Per Child