On Mon, 2008-03-17 at 21:32 -0700, Arjan van de Ven wrote:
> Kevin Hunter wrote:
> > 
> > The more general, more on-topic issue is how to make sleep "just
> > work".  That would save power in a less esoteric way than, say, the
> > tickless kernel.
> 
> there's a few big steps forward coming soon; specifically, the graphic drivers
> are partially moving into the kernel. This allows for a not-insane way of 
> doing
> suspend/resume for these, at least for those folks using the open drivers.
> 
> Video is the big showstopper for suspend/resume right now; not only does it go
> wrong in graphics too often, not having graphics resumed also means you can't
> see printk's and other diagnostics, so other issues remain very well hidden ;(
> With this getting fixed in the next few months, I'm hoping there'll be a big 
> jump
> in quality for S&R on Linux.

How will this work?
Currently nearly all new laptops (even desktops) do suspend to ram
properly. The main problem is to detect whether the BIOS needs to be
posted. With an easy userspace workaround you get s2ram working on >90%
of recent laptops by trying out whether video posting is needed or not,
e.g. with these tools: http://sourceforge.net/projects/suspend

It will get better, probably means the new in-kernel graphics things
popping up soon?
Will this solution:
  A) Re-setup graphics settings so that storing (on suspend) and
     re-writing (on resume) graphics cards parts are not needed at
     all anymore (this is what "graphics cards posting" does, right?)
  B) Be able to detect whether posting is needed and do it

In case of A, the solution will be very hard to backport as huge driver
changes in kernel are needed (mode setting, etc.)
In case of B, this could be a little part of the in-kernel graphics
changes which could easily be backported, the s2ram tools can be
notified or the posting could be implemented in kernel...

Thanks,

   Thomas

_______________________________________________
Discuss mailing list
[email protected]
http://mail.lesswatts.org/mailman/listinfo/discuss

Reply via email to