[Bug 1783954] [NEW] [NVIDIA] GPU lockup (screen freeze) after resume from hibernation / suspend

[YS1]

Public bug reported:

 This bug is reported against xorg as per
<https://wiki.ubuntu.com/X/Troubleshooting/Freeze>. It is also reported
to NVIDIA under
<https://devtalk.nvidia.com/default/topic/1037911/linux/-396-45-gpu-
lockup-screen-freeze-after-resume-from-hibernation-suspend/>.

I have been experiencing GPU lockups after resuming from hibernation on
KUbuntu 18.04.

When this happens, the display freezes except for the mouse cursor. It
is possible to ssh into the computer but killing X does not update the
display either.

These lockups seem to happen at a random time after hibernation. Sometimes the computer works correctly for hours after resuming, sometimes the lockup happen a few seconds after resuming. Actions that are prone to trigger the lockup are :
 - hovering over some task in KDE/Plasma panel and causing the miniature of the application window to be displayed (when a lockup happens, it is often when I do this, but doing this does not systematically incur a lockup) ;
 - with Spectacle (screenshot utility), start selecting a rectangular region (this has systematically caused a lockup when tried after resuming from hibernation/suspend).

Following <https://wiki.ubuntu.com/X/Troubleshooting/Freeze> I tried
with the latest graphics drivers from the graphics-drivers PPA, the
latest version of Xorg in the xorg-edgers PPA and the latest available
kernel (4.17.0-996-generic).

Attached are some logs, including a partial log produced by nvidia-bug-report. Indeed, the script could not complete. Inspection of the running processes with a second ssh connection while the script was hanging up shows that the blocking point is [code]cat /proc/driver/nvidia/./gpus/0000:01:00.0/information[/code]. Indeed, trying to manually read that file in the ssh console also caused it to hang. Of course doing this in normal operation works and gives
[code]Model:           Quadro K620
IRQ:             127
GPU UUID:        GPU-ea026ffc-2c0d-1301-c365-b1cceeb45601
Video BIOS:      82.07.7a.00.1a
Bus Type:        PCIe
DMA Size:        40 bits
DMA Mask:        0xffffffffff
Bus Location:    0000:01:00.0
Device Minor:    0[/code]

** Affects: xorg-server (Ubuntu)
     Importance: Undecided
         Status: Incomplete


** Tags: nvidia

** Attachment added: "dmesg.log"
   https://bugs.launchpad.net/bugs/1783954/+attachment/5168472/+files/dmesg.log

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to grub2 in Ubuntu.
https://bugs.launchpad.net/bugs/1783954

Title:
  [NVIDIA] GPU lockup (screen freeze) after resume from hibernation /
  suspend

Status in xorg-server package in Ubuntu:
  Incomplete

Bug description:
   This bug is reported against xorg as per
  <https://wiki.ubuntu.com/X/Troubleshooting/Freeze>. It is also
  reported to NVIDIA under
  <https://devtalk.nvidia.com/default/topic/1037911/linux/-396-45-gpu-
  lockup-screen-freeze-after-resume-from-hibernation-suspend/>.

  I have been experiencing GPU lockups after resuming from hibernation
  on KUbuntu 18.04.

  When this happens, the display freezes except for the mouse cursor. It
  is possible to ssh into the computer but killing X does not update the
  display either.

  These lockups seem to happen at a random time after hibernation. Sometimes the computer works correctly for hours after resuming, sometimes the lockup happen a few seconds after resuming. Actions that are prone to trigger the lockup are :
   - hovering over some task in KDE/Plasma panel and causing the miniature of the application window to be displayed (when a lockup happens, it is often when I do this, but doing this does not systematically incur a lockup) ;
   - with Spectacle (screenshot utility), start selecting a rectangular region (this has systematically caused a lockup when tried after resuming from hibernation/suspend).

  Following <https://wiki.ubuntu.com/X/Troubleshooting/Freeze> I tried
  with the latest graphics drivers from the graphics-drivers PPA, the
  latest version of Xorg in the xorg-edgers PPA and the latest available
  kernel (4.17.0-996-generic).

  Attached are some logs, including a partial log produced by nvidia-bug-report. Indeed, the script could not complete. Inspection of the running processes with a second ssh connection while the script was hanging up shows that the blocking point is [code]cat /proc/driver/nvidia/./gpus/0000:01:00.0/information[/code]. Indeed, trying to manually read that file in the ssh console also caused it to hang. Of course doing this in normal operation works and gives
  [code]Model:           Quadro K620
  IRQ:             127
  GPU UUID:        GPU-ea026ffc-2c0d-1301-c365-b1cceeb45601
  Video BIOS:      82.07.7a.00.1a
  Bus Type:        PCIe
  DMA Size:        40 bits
  DMA Mask:        0xffffffffff
  Bus Location:    0000:01:00.0
  Device Minor:    0[/code]

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/xorg-server/+bug/1783954/+subscriptions