29x: suspend works fine, resume not

[idl0r]

Hey everybody,

over the past months I tested a few different kernels as well as nvidia-drivers but I don't get a working resume, it always ends up in nvidia-drivers causing trouble.

Brief summary:
I have a ION2 system (Zotac IONITX-U) with VDR (Video Disc Recorder) and Xorg (no DM/WM so no KDE, gnome etc.) with (currently) kernel-3.2.9, nvidia-drivers 295.20 and using pm-utils for suspend/resume.

pm-utils is configured to unload *all* modules during suspend and load them again on resume. That works fine so far until I start the X server again (after resume).
X will be started via "/usr/bin/X -nolisten tcp :0 -quiet -br -nocursor".

Running hook /usr/lib64/pm-utils/sleep.d/75modules suspend suspend:
Unloading kernel module dvb_core...Done.
Unloading kernel module stv090x...Done.
Unloading kernel module lnbp21...Done.
Unloading kernel module stv6110x...Done.
Unloading kernel module tda18212dd...Done.
Unloading kernel module stv0367dd...Done.
Unloading kernel module ddbridge...Done.
Unloading kernel module soundcore...Done.
Unloading kernel module rc_core...Done.
Unloading kernel module snd...Done.
Unloading kernel module nvidia...Done.

# lsmod
Module                  Size  Used by

So on resume the nvidia module will be loaded again (see nvidia-X11.log) and as I said, that's fine until I start/use X again.

/var/log/kern.log:Mar  3 03:32:55 vdr kernel: [  224.978706] NVRM: Xid (0000:04:00): 57, Failed switching to lower mclk freq. in sddr3 link training
/var/log/kern.log:Mar  3 03:36:16 vdr kernel: [  425.910789] NVRM: Xid (0000:04:00): 57, Failed switching to lower mclk freq. in sddr3 link training

(see nvidia-X11.log for more errors)

That causes X11 to hang and only kill -9 works (even that takes some time till X really terminated) but just restarting X doesn't help either.
The only "solution" is to reboot the system. So there's no output X only hangs and even the console (tty) doesn't appear after resume.
I also tried to enable the BIOS feature that re-init the VGA BIOS on post S3 but that didn't help either.
As you can see in the log, it also seems that alsa is having issues to initialize the HDMI stuff again but I don't think *that* is causing all the trouble.

The same happened on my previous system (AMD CPU) with an GT210 PCIe.

pm-utils config:

HOOK_BLACKLIST="00powersave 01grub 49bluetooth 55NetworkManager 90clock 95led 98video-quirk-db-handler 99video"
SUSPEND_MODULES="dvb_core stv090x lnbp21 stv6110x tda18212dd stv0367dd ddbridge soundcore rc_core snd nvidia"

I don't know yet if the alsa errors are just side effects or separate issue but that seems in *this* case not to be related to me.

Any ideas how to solve or at least workaround the issue? Can anybody else confirm that?

[idl0r]

*push

Just out of curiousity, why exactly do you unload all modules? Isn't the usual approach to just unload the modules, which are not needed? Suspend/Resume works most of the time even with nvidia driver loaded for me on a different system...

[idl0r]

Originally Posted by luother

Just out of curiousity, why exactly do you unload all modules? Isn't the usual approach to just unload the modules, which are not needed? Suspend/Resume works most of the time even with nvidia driver loaded for me on a different system...

It doesn't matter if the modules are unloaded or not - it doesn't work, same error so far.

[ 570.754732] NVRM: Xid (0000:04:00): 57, Failed switching to lower mclk freq. in sddr3 link training
[ 571.083178] NVRM: Xid (0000:04:00): 56, CMDre 00000000 00000000 00000000 00000001 00000001

I unload all those modules because:
a) I have one driver which doesn't work when not unloaded before but it works when I unload/load it.
b) There are also some issues with alsa (hda_intel and probably nvidia(HDMI) related) and I thought it would help there as well.