suspend/restore almost works on Sony PCG-GRX570, acpi_video not good

[George Hartzell]

George Hartzell writes:
 > Ian Dowse writes:
 >  > In message <17077.62001.211846.349890 at satchel.alerce.com>, George Hartzell writ
 >  > es:
 >  > >So, I'm trying to figure out what to try next.
 >  > >
 >  > >Can someone suggest something to try?
 >  > 
 >  > One more thing you could try is `kldload vesa' before suspending -
 >  > there is some code in the VESA driver that asks the BIOS to restore
 >  > the video state when the system resumes, which might help.
 > 
 > Ok, I've tried kloading vesa with and without hw.acpi.reset_video and
 > both cases make matters worse, I never see the disk activity that
 > suggests that the system is limping along, and the power button
 > doesn't try to do a shutdown.
 > [...]

I tried this problem on freebsd-current, but didn't catch any
interest, and am hoping that FreeBSD-mobile is more germane.

Quick summary, I have a Sony PCG-GRX570 and I'm trying to run -CURRENT
from a couple of days ago.

Various configuration files and output are available at:

   http://grapeape.alerce.com/KONG 

It generally works, but it's having problems with suspend/resume.  So
far, I've been working from the console.  With hw.acpi.reset_video=1
and hw.acpi.sleep_delay=4, it dims the screen, spings things down, and
starts the red-I'm-asleep LED flashing.  When I resume the screen
brightens and the disk exhibits normal-ish behaviour.  I can type
blindly and things seem to happen (e.g. shutdown -p now causes the
disk to spin for a while) but it never seems to make it all of the way
down.  But, the filesystems are clean when it boots after I hold the
power button for a while.  I can also quickly hit the power button and
seem to get the same attempt to shutdown.  /var/log/messages contains
a message about resuming.

I can suspend and resume from Ubuntu 5.04 Linux live-cd from a text
console, X comes back SNAFU.  Ubuntu also only dims the screen,
doesn't seem to turn off the backlight.

I understand that the DPMS hack to acpi_video is one way to handle the
backlight.  acpi_video doesn't attach since my bios doesn't provide an
acpi-way to twiddle the brightness.  Out of curiousity I hacked it up
to attach anyway [sadly, it attached to 10 devices...] and got the
backlight to go out.  That's a hurdle that I can save for later. 

I'm working on the assumption that there's something about the way
that the console and/or video drivers are being handled that isn't
quite right.  I'm hoping that I can compare and contrast with the way
that Linux is doing it.

I'm also looking at trying to use dcons and dconschat to hook up a
firewire console and see if I can get anywhere with that.

I'd appreciate _any_ guidance on interesting places to start my
comparison.

Thanks,

g.