On Wed, 2005-01-26 at 01:31 +0000, Kentaro A. Kurahone wrote:
> On Tue, Jan 25, 2005 at 09:19:34PM -0400, Jared D. McNeill wrote:
> > On Tue, 2005-01-25 at 17:00 -0800, Jason Thorpe wrote:
> > > On Jan 25, 2005, at 4:18 PM, Matt Thomas wrote:
> > > 
> > > > Well, you almost want the driver to soft-detach, and then go
> > > > through autoconf again on power-on giving preferences to those
> > > > things that soft-detached.
> > > 
> > > That really kind of depends on what sort of bus you're talking about, 
> > > but yah, that's also true.
> > > 
> > > (You certainly want to do that with USB, 1394, etc.)
> > 
> > What about the case of a umass device where we have a filesystem mounted
> > from it? Should we unmount before sleeping, or handle this device
> > differently? (I ask because I'm looking into a suitable powerhook for
> > piixide in ACPI S3 sleep right now, and I can't very well unmount the
> > root filesystem before suspending).
> 
> For piixide, you shouldn't need to.  Just saveoff/restore the pci config
> space, and thwack the controller.

Here's a preliminary patch that actually makes my laptop recover from S3
sleep (first time I've ever seen this happen with NetBSD!).

  http://www.invisible.ca/~jmcneill/netbsd/d600/acpi-sleep-piixide.patch

It doesn't of course ensure that all data is sync'd to disk, but it's a
HUGE improvement over the behaviour I was seeing before. Combined with
my acpi-wakecode patch, I'm actually able to drop to S3 and resume with
my Latitude D600 (still missing some powerhooks for bge and cbb, but
I'll work on that next).

If there aren't any objections, I'd like to commit both the
acpi-sleep-piixide and acpi-wakecode patches to HEAD.

Cheers,
Jared