Re: [TuxOnIce-devel] Why does ToI S-2-Ram resume OK but not …

Hi.

On Sat, 2008-06-07 at 15:57 -0700, Kenneth Crudup wrote:
> I have an HP Pavilion dv9000; it's got a dual-core 64-bit T9300 and an
> NVidia 8600M GS (which, surprisingly enough, works flawlessly; both
> pm-utils hibernate and ToI hibernates, compiz works, the whole thing.
> I was incredulous at first :).
>
> I'm running my own kernels (up to 2.6.25.5 now, no problems so far), too.
>
> What I've done to maximize compatibility with the "Ubuntu Method", as
> well as trying not to break any of the automatic updates (I'm learning
> to like package managers again, which is really tough for a "download/
> configure/patch/build/install" kind of guy like me) is just change the
> final calls in /usr/lib/pm-utils/functions to use hibernate instead of
> echo "<method>" > /sys/power/state. This works OK for replacing
> hibernate (I realize this method means two sets of scripts run, but it
> seems less intrusive to whatever they might add later on).
>
> I can also suspend-to-ram, but not with save-to-both in the mix. I
> can replace "echo"s into the appropriate /sys/power/... files with
> "hibernate-ram", and that does suspend-to-ram, but if I then prepend
>
> "echo 3 > /sys/power/tuxonice/suspend_method"
>
> before that hibernate call, it writes to disk then suspends OK, but on
> the way back from suspend the system *is* running, but I end up with the
> dreaded "NVidia Black Screen of Resume" where the backlight is off and I
> don't think the keys are working, but the SysRq stuff is still responding
> ('cause that's how I sync and r/o-mount then reboot).
>
> I looked thru the lists.tuxonice.net archives and didn't really catch a
> whole lot of input to this; has anyone figured out what's going on?
> The dmesg and /var/log/syslog files were intact, but don't show any
> issues (and there's nothing out of the NVidia driver save the normal
> "NVRM: RmPowerManagement:" messages).
>
> Anyone got any ideas to try? I don't normally have to have "switch to
> text mode" turned on with this setup 'cause VT switching on this version
> of the NVidia driver takes so long (but the pm-utils stuff chvt's to tty63
> anyway), but I'd tried that and added the appropriate hack to try and
> restore the DPMS while in a VT; no dice.

I have this too, and want to fix it, but am struggling - as usual! - to
find the time. I assume I'm doing something wrong with device
suspend/resume calls and/or ordering of calls.

Regards,

Nigel