WARNING for NVIDA users: kernel 5.9 coming in sid

Begonnen von Camelot, 2020/11/02, 11:02:58

Vorheriges Thema - Nächstes Thema

Camelot

As of today, sid/unstable comes with kernel 5.9, NVIDIA users should hold back kernel 5.8.×.

apt-mark hold linux-headers-siduction-amd64
apt-mark hold linux-image-siduction-amd64


The following packages have been kept back:
   linux-headers-siduction-amd64 (5.8-17 => 5.9-3)
   linux-image-siduction-amd64 (5.8-17 => 5.9-3)
0 upgraded, 0 newly installed, 0 to remove and 2 not upgraded.



Why?: NVIDIA Doesn't Expect To Have Linux 5.9 Driver Support For Another Month

But don't forget, if NVIDIA drivers work again...

der_bud

Du lachst? Wieso lachst du? Das ist doch oft so, Leute lachen erst und dann sind sie tot.

sunrat

One of the admins at Debian User Forums reported the current 450.80.02-1 nvidia driver in sid works with the current sid 5.9 kernel - http://forums.debian.net/viewtopic.php?f=30&t=147803#p729675
I queried him about it as I read that Phoronix article too. I haven't tested it as I use Nouveau on my siduction system.

Penyelam

I've experimental in the repos, it works:

$ inxi -SG
System:    Host: siductionbox Kernel: 5.9.3-towo.1-siduction-amd64 x86_64 bits: 64 Console: tty 1
           Distro: siduction 17.1.0 Patience - kde - (201703051755)
Graphics:  Device-1: NVIDIA GK208B [GeForce GT 730] driver: nvidia v: 450.80.02
           Display: server: X.org 1.20.9 driver: nvidia unloaded: fbdev,modesetting,nouveau,vesa tty: 189x52
           Message: Advanced graphics data unavailable in console. Try -G --display

terroreek

I can confirm the nvidia driver in unstable works with the 5.9.3 kernel


manilg

Same here:

inxi -v1
System:    Host: rhea Kernel: 5.9.3-towo.1-siduction-amd64 x86_64 bits: 64 Desktop: LXQt 0.14.1
           Distro: siduction 18.3.0 Patience - lxqt - (201805132142)
CPU:       Info: Quad Core Intel Core i7-4790 [MT MCP] speed: 3600 MHz min/max: 800/3600 MHz
Graphics:  Device-1: NVIDIA GK106 [GeForce GTX 650 Ti] driver: nvidia v: 450.80.02
           Display: x11 server: X.Org 1.20.9 driver: nvidia resolution:


The driver was build quckly, but afterwords it needed 5 minutes till the installation of the new kernel was finished.
After reboot, all is fine.

manilg

seasons

The main nvidia module will work with kernel 5.9. However, the nvidia-uvm module will not. The upshot is that those that need CUDA and CUDA-based functions (like NVDEC/NVENC) should keep a 5.8 kernel handy until Nvidia releases a driver that works around the GPL symbol issue.

Camelot

I confirm the driver 450.80.02 works with kernel 5.9.3.
System:    Host: sidzeug Kernel: 5.9.3-towo.1-siduction-amd64 x86_64 bits: 64 Desktop: Xfce 4.14.2
           Distro: siduction 18.3.0 Patience - xfce - (201805132203)
Graphics:  Device-1: NVIDIA GK208B [GeForce GT 730] driver: nvidia v: 450.80.02
           Display: x11 server: X.Org 1.20.9 driver: nvidia resolution: 1920x1080~60Hz
           OpenGL: renderer: GeForce GT 730/PCIe/SSE2 v: 4.6.0 NVIDIA 450.80.02


Normal reboot without warning or so.

ajavibp

linux-image-5.9.3-towo.1-siduction-amd64 working perfectly with the nvidia legacy 340 drivers.
This was not what I expected!  ;D

seasons

Zitat von: ajavibp in 2020/11/03, 18:01:33
linux-image-5.9.3-towo.1-siduction-amd64 working perfectly with the nvidia legacy 340 drivers.
This was not what I expected!  ;D
Why not? There is nothing special about kernel 5.9 and the main part of the nvidia driver. As usual, patches have been floating out there since the 5.9 rc's and the Debian maintainers applied them: (For 340 driver, see: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=972430 )

The symbol issue affects the nvidia-uvm module, which is the Nvidia Unified Memory module, used for CUDA.

ajavibp

Legacy drivers tend to stop working with each new version, and sometimes it takes a long time to work again.

dibl

Nvidia driver 455.38 from experimental is installing and working correctly on kernel 5.9.3-towo.2 (2 GPUs):

System:    Host: dibl-patience Kernel: 5.9.3-towo.2-siduction-amd64 x86_64 bits: 64 Desktop: KDE Plasma 5.19.5
           Distro: siduction 18.3.0 Patience - kde - (201805132121)
CPU:       Info: Quad Core Intel Core i7-7740X [MT MCP] speed: 4500 MHz min/max: 800/4500 MHz
Graphics:  Device-1: NVIDIA GP106 [GeForce GTX 1060 6GB] driver: nvidia v: 455.38
           Display: x11 server: X.Org 1.20.9 driver: nvidia resolution: 1: 1920x1200~60Hz 2: 1920x1080~60Hz
           OpenGL: renderer: GeForce GTX 1060 6GB/PCIe/SSE2 v: 4.6.0 NVIDIA 455.38
Drives:    Local Storage: total: 2.27 TiB used: 1.01 TiB (44.4%)
Info:      Processes: 277 Uptime: 1m Memory: 31.28 GiB used: 1.65 GiB (5.3%) Shell: Bash inxi: 3.1.08


System76 Oryx Pro, Intel Core i7-11800H, ASRock B860 Pro-A, Intel Core Ultra 7 265KF, Nvidia GTX-1060, SSD 990 EVO Plus.

seasons

Zitat von: dibl in 2020/11/05, 10:47:16
Nvidia driver 455.38 from experimental is installing and working correctly on kernel 5.9.3-towo.2 (2 GPUs):
So the nvidia-uvm module is loaded and CUDA or NVDEC is working correctly? That's the issue here; not whether the main part of driver installs and works (Read the Article!). We don't need 50 more "works for me" posts if people aren't going to test CUDA.

dibl

Zitat von: seasons...We don't need 50 more "works for me" posts if people aren't going to test CUDA.

OK, how does one test CUDA?  (I have never done a CUDA test).

Thanks.
System76 Oryx Pro, Intel Core i7-11800H, ASRock B860 Pro-A, Intel Core Ultra 7 265KF, Nvidia GTX-1060, SSD 990 EVO Plus.

seasons

Probably the easiest way to check CUDA is by checking if nvidia-uvm is loaded:
lsmod | grep nvidia