[EN] [SOLVED] Nvidia graphics, xorg memory leak since kernel 4.19.2

[dibl]

I have 4 siduction systems, only 1 has Nvidia graphics -- the others are all Intel graphics.  The one with Nvidia graphics is using the nvidia-driver.

On the system with Nvidia graphics, I have been seeing a serious memory leak since at least the 4.19.2 kernel -- possibly it started earlier.  The culprit is xorg.  The platform has 32 G of memory, and if I let it run overnight, in the morning it will be starting to swap, and the problem process is xorg.

Does anyone else see this issue?

[piper]

Which nvidia driver are you using?

I cannot reproduce this with experimental 396.54-2

The only thing I can find is

https://devtalk.nvidia.com/default/topic/1043938/nvidia-geforce-4xx-series-drivers-segfault-with-kernel-4-19-x/

But not related here (driver wise)

[dibl]

Code: [Select]

don@dibl-patience:~$ inxi -G
Graphics:
  Device-1: NVIDIA GP106 [GeForce GTX 1060 6GB] driver: nvidia v: 390.87
  Display: x11 server: X.Org 1.20.3 driver: nvidia
  resolution: 1920x1200~60Hz, 1920x1080~60Hz
  OpenGL: renderer: llvmpipe (LLVM 7.0 256 bits) v: 3.3 Mesa 18.2.6

Is that v. 396 driver in the experimental repo?

[dibl]

Here's a picture.

[piper]

Yes 396.54-2 is in experimental

[piper]

Code: [Select]

07:29:04 PM-Fri Nov 30-[piper@x1] $ inxi -G
Graphics:  Device-1: NVIDIA GM204 [GeForce GTX 970] driver: nvidia v: 396.54
           Display: x11 server: X.Org 1.20.3 driver: nvidia unloaded: fbdev,modesetting,nouveau,vesa
           resolution: 1920x1080~60Hz
           OpenGL: renderer: GeForce GTX 970/PCIe/SSE2 v: 4.6.0 NVIDIA 396.54

[dibl]

OK, fully updated, Torsten's latest kernel installed, and the 396 Nvidia driver from experimental.

Code: [Select]

System:
  Host: dibl-patience Kernel: 4.19.5-towo.3-siduction-amd64 x86_64 bits: 64
  Desktop: KDE Plasma 5.14.3 Distro: siduction 18.3.0 Patience - kde - (201805132121)
Machine:
  Type: Desktop Mobo: ASUSTeK model: ROG STRIX X299-E GAMING v: Rev 1.xx
  serial: <filter> UEFI: American Megatrends v: 1401 date: 05/21/2018
CPU:
  Topology: Quad Core model: Intel Core i7-7740X bits: 64 type: MT MCP
  L2 cache: 8192 KiB
  Speed: 3668 MHz min/max: 800/4500 MHz Core speeds (MHz): 1: 800 2: 800 3: 800 4: 800
  5: 800 6: 800 7: 800 8: 800
Graphics:
  Device-1: NVIDIA GP106 [GeForce GTX 1060 6GB] driver: nvidia v: 396.54
  Display: x11 server: X.Org 1.20.3 driver: nvidia
  resolution: 1920x1200~60Hz, 1920x1080~60Hz
  OpenGL: renderer: llvmpipe (LLVM 7.0 256 bits) v: 3.3 Mesa 18.2.6
Audio:
  Device-1: Intel 200 Series PCH HD Audio driver: snd_hda_intel
  Device-2: NVIDIA GP106 High Definition Audio driver: snd_hda_intel
  Device-3: Logitech HD Pro Webcam C920 type: USB driver: snd-usb-audio,uvcvideo
  Sound Server: ALSA v: k4.19.5-towo.3-siduction-amd64
Network:
  Device-1: Intel Ethernet I219-V driver: e1000e
  IF: enp0s31f6 state: up speed: 1000 Mbps duplex: full mac: <filter>
  IF-ID-1: br0 state: up speed: N/A duplex: N/A mac: <filter>
Drives:
  Local Storage: total: 2.27 TiB used: 910.28 GiB (39.1%)
  ID-1: /dev/sda vendor: Samsung model: SSD 860 EVO 500GB size: 465.76 GiB
  ID-2: /dev/sdb vendor: Western Digital model: WD1003FZEX-00K3CA0 size: 931.51 GiB
  ID-3: /dev/sdc vendor: Western Digital model: WD1003FZEX-00K3CA0 size: 931.51 GiB
Partition:
  ID-1: / size: 58.81 GiB used: 10.19 GiB (17.3%) fs: ext4 dev: /dev/sda2
  ID-2: /home size: 393.18 GiB used: 51.12 GiB (13.0%) fs: ext4 dev: /dev/sda4
  ID-3: swap-1 size: 5.00 GiB used: 0 KiB (0.0%) fs: swap dev: /dev/sda3
Sensors:
  System Temperatures: cpu: 35.0 C mobo: N/A gpu: nvidia temp: 33 C
  Fan Speeds (RPM): cpu: 0 gpu: nvidia fan: 0%
Info:
  Processes: 265 Uptime: 2m Memory: 31.33 GiB used: 1.72 GiB (5.5%) Shell: bash
  inxi: 3.0.27
I will let it run overnight and see what I have in the morning.  Thanks, piper!

[piper]

No problems dibl, good luck

[towo]

Code: [Select]

driver: nvidia v: 396.54
OpenGL: renderer: llvmpipe (LLVM 7.0 256 bits) v: 3.3 Mesa 18.2.6
You see the problem?

[piper]

I'm wondering if he has a

Code: [Select]

nvidia-blacklists-nouveau.conf
in

Code: [Select]

/etc/modprobe.d

I always just do a

Code: [Select]

apt purge xserver-xorg-video-nouveau
But, I don't recommend that to newer users

[dibl]

Thanks towo -- I was a little too fast with the driver from experimental, and forgot the mesa packages.  Does this look better?

Code: [Select]

Graphics:
  Device-1: NVIDIA GP106 [GeForce GTX 1060 6GB] driver: nvidia v: 396.54
  Display: x11 server: X.Org 1.20.3 driver: nvidia
  resolution: 1920x1200~60Hz, 1920x1080~60Hz
  OpenGL: renderer: llvmpipe (LLVM 7.0 256 bits) v: 3.3 Mesa 18.3.0-rc5

Code: [Select]

don@dibl-patience:/$ glxinfo | grep OpenGL
OpenGL vendor string: VMware, Inc.
OpenGL renderer string: llvmpipe (LLVM 7.0, 256 bits)
OpenGL core profile version string: 3.3 (Core Profile) Mesa 18.3.0-rc5
OpenGL core profile shading language version string: 3.30
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile
OpenGL core profile extensions:
OpenGL version string: 3.1 Mesa 18.3.0-rc5
OpenGL shading language version string: 1.40
OpenGL context flags: (none)
OpenGL extensions:
OpenGL ES profile version string: OpenGL ES 3.0 Mesa 18.3.0-rc5
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.00
OpenGL ES profile extensions:

I do have the nouveau driver blacklisted and the package is not installed -- thanks piper.  Now that I fixed the OpenGLX version situation, it will take some time to see if the memory issue is fixed.

[piper]

I question the "blacklist thing" only because what I read at opensuse to a few people

[towo]

Code: [Select]

OpenGL: renderer: llvmpipe (LLVM 7.0 256 bits) v: 3.3 Mesa 18.3.0-rc5No, that's even wrong!
Your path to the right glx lib seems to be wrong. After a look in the changelog of the nvidia driver, there was a packaging error which sould be fixed in 296.87-3

Revert move of libglx.so to /usr/lib/nvidia/current/modules.

I switched to ATI, so i can't investigate fourther.

[piper]

What I would do first

Code: [Select]

apt purge nvidia.
Yes, use the period  .

Code: [Select]

apt update; apt install nvidia-driver nvidia-driver-libs nvidia-settings nvidia-smi nvidia-kernel-dkms -t experimental
If you don't need/use thermal monitors you don't have to include nvidia-smi

Post back if this don't work

[dibl]

Great -- thanks guys.  I knew something is wrong -- xorg continues to slowly suck up memory, even when no user packages are running.  I will try these ideas.