v8.0.0: Device crashes when entering suspended mode (nVidia driver issue)

  • Hi all. Great stuff on the newest Beta. Loving everything so far.

    Except one thing. Ever since upgrading from LE 7.0.3 to LE 7.95.1 I'm having power-saving problems. When the device goes to sleep (suspended) or when I manually suspend the device from the menu, it won't wake up again.

    It doesn't respond to:

    • IR remote
    • Keyboard (USB)
    • Wake-on-Lan signal
    • Power button (short press)!


    The only thing I can do is long press the power button to shut it down completely, and turn it on again.

    Hardware:
    Zotac ZBOX ID80
    Intel Atom D2700 CPU (dual-core 2.13GHz)
    Nvidia® GeForce® GT 520M

    Image: LibreELEC-Generic.x86_64-7.95.1.tar

    Debug log: nweptDcS

    As far as I can see, the log doesn't show anything interesting. It appears to be entering suspended mode normally. Did something change between 7.0.3 and LE 7.95.1 that requires a bios settings change?

    Apologies if this has come up before. I couldn't find an answer on the forums and in the github issue tracker.

    Edited once, last by Sir.Swaffel (February 22, 2017 at 9:35 AM).

  • Same here with a Nvidia GTX-1060 6Gb card.

    I blame it on the Nvidia linux driver, the issue is all over the nvidia drivers forum like in this thread: 8

    It should be fixed with the latest driver 378.09 , but obviously some people still have issues.

    Hope Nvidia will fix that soon.

  • Having a htpc with a GT520 myself, going into suspend freezes my machine.

    Try enforcing the legacy Nvidia video driver

    Code
    wget http://chewitt.libreelec.tv/96-nvidia.rules -O /storage/.config/udev.rules.d/96-nvidia.rules

    It has solved the suspend problem on my machine, and so far I haven't noticed any video performance problems. The actual problem is hard to tackle so far.

  • Thanks for both your replies.

    Klojum's trick seems to work. Not sure what the downsides are to using legacy nVidia drivers, but I guess I'll use them until the linux drivers are updated and included in the next LE beta.


  • Hi all. Great stuff on the newest Beta. Loving everything so far.

    Except one thing. Ever since upgrading from LE 7.0.3 to LE 7.95.1 I'm having power-saving problems. When the device goes to sleep (suspended) or when I manually suspend the ...
    ...

    ...
    requires a bios settings change?

    Apologies if this has come up before. I couldn't find an answer on the forums and in the github issue tracker.

    Just to confirm: I'm having the same setup and same problem. Reproducible

    Thanks, will switch to the (temporarily) solution for now.

    Vincent

  • Using LibreElec 7.95.2 with HTPC Geforce 610 makes the System freeze upon sleep.

    Klojums method works for me too :)
    Sleep is working with the enforced legacy drivers.

  • i have the same problem, but it would be fne to fix it from libreelec team byself.

    7.95.2 still have the problem

    and the legacy driver is not good for watching live tv... i will go back to 7.0.3

    Edited once, last by dirki1972 (February 4, 2017 at 7:48 PM).


  • On the nVidia forums users are reporting that the latest Linux drivers seem to fix it.

    So hopefully they will be included in 7.95.3


    7.95.3 still has the same nvidia drivers as the previous betas, so no fixes for our problem. It seems the bug makes it's way into final release 8.0 :(

  • 7.95.3 still has the same nvidia drivers as the previous betas, so no fixes for our problem. It seems the bug makes it's way into final release 8.0 :(

    If so, I guess we'll have to compile it ourselves.

    Might it be as simple as changing: LibreELEC.tv/package.mk at master · LibreELEC/LibreELEC.tv · GitHub

    And running "python packages/x11/driver/xf86-video-nvidia/scripts/make_nvidia_udev.py" before compiling?

    Edited once, last by Sir.Swaffel (February 11, 2017 at 3:18 PM).

  • Same for me (GT 610), Systems hangs when going to suspend and CPU gets very hot while system hangs. Fix worked for me, suspend is working now. THX

    Edited once, last by cyablo (February 12, 2017 at 11:20 AM).

  • Decided to compile the master branch with 378.09 beta drivers, and getting weird results (black screen with a single underscore, nothing in the logs) when going into suspend. So I guess I'll try an older version 367.57 next.

    Edited once, last by Sir.Swaffel (February 12, 2017 at 8:58 PM).


  • Decided to compile the master branch with 378.09 beta drivers, and getting weird results (black screen with a single underscore, nothing in the logs) when going into suspend. So I guess I'll try an older version 367.57 next.

    Thats excactly how v7.95.3 BETA behaved for me without any modification.


  • Do you know which drivers came with LE 7.03? These worked perfect.

    That would be 352.79, which is quite old.

    Anything above 375.xx seems to have this bug, so 367.57 would be the best choice. It's compiling now, will report back with findings.

    Edited once, last by Sir.Swaffel (February 13, 2017 at 1:49 PM).

  • Update: can't compile the master branch with 367.57.

    Placing the kernel patch file from xf86-video-nvidia: linux 4.8 fix · LibreELEC/LibreELEC.tv@af5d3ef · GitHub (and renaming to 4.9) doesn't work, but since I'm not if it's applicable to 4.9, I've got no idea.

    Edited once, last by Sir.Swaffel (February 13, 2017 at 3:50 PM).

  • Seems that the driver has been updated to 375.39 with xf86-video-nvidia: update to xf86-video-nvidia-375.39 · LibreELEC/LibreELEC.tv@2c04648 · GitHub

    The changelog for the new driver:

    Quote


    Release highlights since 375.26:

    • Fixed a bug that caused system hangs when resuming from suspend with some GPUs.
    • Fixed a regression that could cause corruption when hot-plugging displays.
    • Fixed a regression that prevented systems with multiple DisplayPort monitors from resuming correctly from suspend.
    • Fixed a bug that caused hangs when using the NVIDIA driver on 32-bit x86 systems running older Linux kernel versions (e.g. 2.6.18).

    I can't test it right now, but I'll assume this bug is now fixed until reported otherwise.