nvidia-390xx AUR package discussion thread

Filip62 · 2021-07-26 18:07:29

Works fine on 5.13, thanks.

jonathon · 2021-09-04 11:34:26

Updated for 5.14. Module compiles but (as usual) I have no way of testing it.

claudecat · 2021-09-04 21:32:53

Thanks Jonathan! It's nice to know that the 390 driver is in such capable and attentive hands. Haven't tried the new patched-for-5.14 build yet on Arch, but have no doubt that it will work. (Already manually patched and working on Slackware-current here.)

DAC324 · 2021-09-06 10:35:04

Hello,

Looks like the good old Nvidia driver 390.144 is reaching the end of its days.

I have experimented a bit with the 5.14 kernel. It has a new config option

CONFIG_BPF=y

If CONFIG_BPF is enabled, loading the nvidia driver on boot fails:

Sep 04 11:50:01.776227 gerd-desktop kernel: failed to validate module [nvidia] BTF: -22
Sep 04 11:50:02.022670 gerd-desktop kernel: BPF:[99862] TYPEDEF  
Sep 04 11:50:02.022780 gerd-desktop kernel: BPF:type_id=97039
Sep 04 11:50:02.022813 gerd-desktop kernel: BPF: 
Sep 04 11:50:02.022842 gerd-desktop kernel: BPF:Invalid name
Sep 04 11:50:02.022904 gerd-desktop kernel: BPF:

Problem seems to be an invalid (or empty) name entry

Sep 04 11:50:02.022842 gerd-desktop kernel: BPF:Invalid name

however, there does not seem to be anything related in the sources.

At a first glance, this looks a bit similar to the problem described for kernel 5.13 and the crc32 driver but there, it looks like this driver has an empty BPF record whatsoever.
Hence, the suggested patches Part 1/2 and part 2/2 just handle empty BPF records, and they appear to already be included in Kernel 5.14.1.
Is there something I can do about

Sep 04 11:50:02.022842 gerd-desktop kernel: BPF:Invalid name

as well?

Thank you very much in advance for all helpful thoughts,
DAC324

Last edited by DAC324 (2021-09-06 10:36:45)

jonathon · 2021-11-04 11:10:44

Some good news - DKMS module will build against 5.15 without any changes!

DAC324 wrote:

Is there something I can do about
Sep 04 11:50:02.022842 gerd-desktop kernel: BPF:Invalid name
as well?

I don't know how much this is to do with using a non-standard kernel option or how much to do with the Nvidia driver, but if you can find or create a patch for the driver then I can include it.

bwidlar · 2021-11-10 08:16:09

Hi, nvidia-390xx user here.
First all, thanks to jonathon for the support in this aur package.

No nvidia graphics after pacman update, using nvidia 390.144

[    11.515] (II) LoadModule: "nvidia"
[    11.516] (II) Loading /usr/lib/xorg/modules/drivers/nvidia_drv.so
[    11.524] (II) Module nvidia: vendor="NVIDIA Corporation"
[    11.524] 	compiled for 4.0.2, module version = 1.0.0
[    11.524] 	Module class: X.Org Video Driver
[    11.524] ================ WARNING WARNING WARNING WARNING ================
[    11.524] This server has a video driver ABI version of 25.2 that this
driver does not officially support.  Please check
http://www.nvidia.com/ for driver updates or downgrade to an X
server with a supported driver ABI.
[    11.524] =================================================================
[    11.524] (EE) NVIDIA: Use the -ignoreABI option to override this check.
[    11.524] (II) UnloadModule: "nvidia"
[    11.524] (II) Unloading nvidia
[    11.524] (EE) Failed to load module "nvidia" (unknown error, 0)

I will try the -ignoreABI option updating my xorg configuration file in /etc/X11/xorg.conf.d

Section "ServerFlags"
    Option        "IgnoreABI" "1" 
EndSection

Thanks.

Last edited by bwidlar (2021-11-10 09:04:58)

bwidlar · 2021-11-10 08:22:46

It works, all fine now.
If you think this info is irrelevant to the topic, please delete my posts.

And again, thanks to jonathon for your time in this package, a lot of hardware working thanks to it.
Thanks.

sayonalion · 2021-11-10 08:30:05

bwidlar wrote:

It works, all fine now.
If you think this info is irrelevant to the topic, please delete my posts.
And again, thanks to jonathon for your time in this package, a lot of hardware working thanks to it.
Thanks.

How did you resolve the problem?

seth · 2021-11-10 08:51:13

bwidlar wrote:

I will try the -ignoreABI option.

sayonalion · 2021-11-10 08:59:06

Ok, I solved the problem when the X server doesn't start and returns the `NVIDIA: Use the -ignoreABI option to override this check` message the following way:

# /etc/X11/xorg.conf.d/30-nvidia.conf
Section "ServerFlags"
  Option "IgnoreABI" "true"
EndSection

claudecat · 2021-11-10 14:03:15

I had the same issue as a few others here with X not starting after a recent update, and can confirm that adding the previously mentioned section to xorg.conf (what I have) or (presumably) in an appropriately named file in /etc/X11/xorg.conf.d/ works. I was getting worried last night before anyone had posted anything, being the relatively idiotic Linux user that I am. But once again, as is always the case on Arch, someone has solved the problem within hours of it cropping up, even in corner case hardware situations like this. Thanks to everyone involved!

Ammako · 2021-11-10 16:07:56

Should be able to include that xorg config in the package, to automatically apply it for users.

I'm not sure if the config might break things on older versions of xorg-server, but users are expected to be running the latest version of every package anyway.

seth · 2021-11-10 16:12:52

Option "IgnoreABI" "true" will not harm on 1.20, but while it will allow the 1.21 server to start w/ the driver compiled against 1.20, there's no guarantee that the driver won't "misresovle" a symbol and then crashes the server.
nvidia will have to provide an updated driver at some point. hopefully.

jonathon · 2021-11-10 17:21:22

I can easily include the override file, though I don't want it to cause weird issues. On the other hand, Xorg is currently entirely non-functional without that override...

Edit: override workaround now included in nvidia-390xx-utils.

Last edited by jonathon (2021-11-10 17:36:09)

Juggrnaut · 2021-11-13 17:53:14

After adding in the "IgnoreABI" to "/etc/bumblebee/xorg.conf.nvidia" I now receive the following error and it doesn't load:

(WW) NVIDIA: The driver will continue to load, but may behave strangely.

(WW) NVIDIA: This driver was compiled against the X.Org server SDK from commit 4c00369024c9754e670fca37804711fef2980e2f and may not be compatible with the final version of this SDK.

Any thoughts?

Last edited by Juggrnaut (2021-11-13 17:54:08)

V1del · 2021-11-13 18:55:38

That's a warning and not an error and should not be the cause, what else do you get?

Last edited by V1del (2021-11-13 18:55:59)

Juggrnaut · 2021-11-13 19:23:09

This is the relevant output of my /var/log/Xorg.8.log file:

[  7586.775] (II) Module glx: vendor="NVIDIA Corporation"
[  7586.775] 	compiled for 4.0.2, module version = 1.0.0
[  7586.775] 	Module class: X.Org Server Extension
[  7586.775] (II) NVIDIA GLX Module  390.144  Wed Jun  2 23:02:00 UTC 2021
[  7586.775] (II) LoadModule: "nvidia"
[  7586.776] (II) Loading /usr/lib/xorg/modules/drivers/nvidia_drv.so
[  7586.776] (II) Module nvidia: vendor="NVIDIA Corporation"
[  7586.776] 	compiled for 4.0.2, module version = 1.0.0
[  7586.777] 	Module class: X.Org Video Driver
[  7586.777] ================ WARNING WARNING WARNING WARNING ================
[  7586.777] This server has a video driver ABI version of 25.2 that this
driver does not officially support.  Please check
http://www.nvidia.com/ for driver updates or downgrade to an X
server with a supported driver ABI.
[  7586.777] =================================================================
[  7586.777] (WW) NVIDIA: The driver will continue to load, but may behave strangely.
[  7586.777] (WW) NVIDIA: This driver was compiled against the X.Org server SDK from commit 4c00369024c9754e670fca37804711fef2980e2f and may not be compatible with the final version of this SDK.
[  7586.777] (II) LoadModule: "mouse"
[  7586.777] (WW) Warning, couldn't open module mouse
[  7586.777] (EE) Failed to load module "mouse" (module does not exist, 0)
[  7586.777] (II) LoadModule: "kbd"
[  7586.778] (WW) Warning, couldn't open module kbd
[  7586.778] (EE) Failed to load module "kbd" (module does not exist, 0)
[  7586.778] (II) NVIDIA dlloader X Driver  390.144  Wed Jun  2 23:05:01 UTC 2021
[  7586.778] (II) NVIDIA Unified Driver for all Supported NVIDIA GPUs
[  7586.778] (EE) No devices detected.
[  7586.778] (EE) 
Fatal server error:
[  7586.778] (EE) no screens found(EE) 
[  7586.778] (EE) 
Please consult the The X.Org Foundation support 
	 at http://wiki.x.org
 for help. 
[  7586.778] (EE) Please also check the log file at "/var/log/Xorg.8.log" for additional information.
[  7586.778] (EE) 
[  7586.778] (EE) Server terminated with error (1). Closing log file.

When I try to run "primusrun glxgears", I get:

primus: fatal: Bumblebee daemon reported: error: [XORG] (EE) No devices detected.

seth · 2021-11-13 20:29:17

[  7586.778] (EE) No devices detected.

Open a new thread.
Therer post the entire log (not what you deem relevant, we might deem differently…) as well as the output of "modinfo nvidia | head -10" and "sudo modprobe nvidia" and your system journal ("sudo journalctl -b")

Juggrnaut · 2021-11-14 02:09:03

seth wrote:

[  7586.778] (EE) No devices detected.
Open a new thread.
Therer post the entire log (not what you deem relevant, we might deem differently…) as well as the output of "modinfo nvidia | head -10" and "sudo modprobe nvidia" and your system journal ("sudo journalctl -b")

Thanks for the info. Luckily I've solved my issue and am posting it here in case anyone else runs into the same problem.

Prior to the recent updates, I never needed to modify the "/etc/bumblebee/xorg.conf.nvidia" file and it worked just fine OOTB. Now after the recent updates, I had to modify that file to now also include the BusID of the nvidia card. Details found here: https://wiki.archlinux.org/title/Bumble … IA_Optimus

So after adding in the BusID (along with the "IgnoreABI" setting), it is working again.

Here is my "/etc/bumblebee/xorg.conf.nvidia" file now:

Section "ServerLayout"
    Identifier  "Layout0"
    Option      "AutoAddDevices" "false"
    Option      "AutoAddGPU" "false"
EndSection

Section "ServerFlags"
    Option      "IgnoreABI" "true"
EndSection

Section "Device"
    Identifier  "DiscreteNvidia"
    Driver      "nvidia"
    VendorName  "NVIDIA Corporation"
    BusID       "PCI:01:00:0"

    Option "ProbeAllGpus" "false"

    Option "NoLogo" "true"
    Option "UseEDID" "false"
    Option "UseDisplayDevice" "none"
EndSection

seth · 2021-11-14 07:48:55

https://wiki.archlinux.org/title/NVIDIA#Early_loading ?

V1del · 2021-11-14 12:32:14

Early loading shouldn't be the issue here as it will be loaded on demand anyway/ a new xorg server started.

If that is indeed the content, then there was nothing actually identifying the correct card to use without the BusID and this only worked by happenstance. I would've thought bumblebee config used OutputClass with a MatchDriver or so for proper identification.

dybdeskarphet · 2021-11-17 18:49:37

firstly, i want to thank you all for your work but secondly, i can't make it work on GeForce 610M. i tried both linux-lts and linux kernels. added the option for ignoring ABI but it only detects Intel nevertheless. strange thing is that X still launches but compositor doesn't work and xrandr --listproviders only shows Intel. anyone having this problem?

also, in journalctl, only nvidia related output is:

Kas 17 21:24:45 K55VD kernel: nvidia: loading out-of-tree module taints kernel.
Kas 17 21:24:45 K55VD kernel: nvidia: module license 'NVIDIA' taints kernel.
Kas 17 21:24:45 K55VD kernel: Disabling lock debugging due to kernel taint
Kas 17 21:24:45 K55VD kernel: nvidia: module verification failed: signature and/or required key missing - tainting kernel
Kas 17 21:24:45 K55VD kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 235
Kas 17 21:24:45 K55VD kernel: nvidia 0000:01:00.0: enabling device (0000 -> 0003)
Kas 17 21:24:45 K55VD kernel: nvidia 0000:01:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=none:owns=none
Kas 17 21:24:45 K55VD kernel: NVRM: loading NVIDIA UNIX x86_64 Kernel Module  390.144  Wed Jun  2 23:02:41 UTC 2021 (using threaded interrupts)

and Xorg has this (even after adding the option for ignoring ABI):

[     8.163] ================ WARNING WARNING WARNING WARNING ================
[     8.163] This server has a video driver ABI version of 25.2 that this
driver does not officially support.  Please check
http://www.nvidia.com/ for driver updates or downgrade to an X
server with a supported driver ABI.
[     8.163] =================================================================
[     8.163] (WW) NVIDIA: The driver will continue to load, but may behave strangely.
[     8.163] (WW) NVIDIA: This driver was compiled against the X.Org server SDK from commit 4c00369024c9754e670fca37804711fef2980e2f and may not be compatible with the final version of this SDK.

Last edited by dybdeskarphet (2021-11-17 18:58:56)

seth · 2021-11-17 21:05:04

Please post the entire xorg log - the excerpt is useless except that it actually suggests that the nvidia GPU is being detected…

dybdeskarphet · 2021-11-18 08:44:17

alright, here it is: https://0bin.net/paste/lsm946jg#BSsaptM … f5z+LRjbJq

also looks like i missed one error, there is also this:

[     8.210] (II) Initializing extension DRI2
[     8.210] (II) Initializing extension GLX
[     8.211] (EE) Failed to initialize GLX extension (Compatible NVIDIA X driver not found)
[     8.216] (II) intel(0): switch to mode 1366x768@60.0 on LVDS1 using pipe 0, position (0, 0), rotation normal, reflection none
[     8.226] (II) intel(0): Setting screen physical size to 361 x 203

seth · 2021-11-18 09:24:05

Looks misconfigued - the system runs on the intel driver but loads the nvidia glx module because that path is preferred in the ModulePath, what's more applicable to an nvidia-only setup.

ls -R /{etc,usr/share}/X11/xorg.conf*

and see https://wiki.archlinux.org/title/NVIDIA_Optimus
FIgure how you want to use the system and configure it accordingly. In doubt create a new thread about it.

Edit: ffwi, partition uuids and memory positions of pci devices are not sensitive data in any way shape or form (the only purpose of the uuid is to avoid collisions)

Last edited by seth (2021-11-18 09:25:08)

Arch Linux

#51 2021-07-26 18:07:29

Re: nvidia-390xx AUR package discussion thread

#52 2021-09-04 11:34:26

Re: nvidia-390xx AUR package discussion thread

#53 2021-09-04 21:32:53

Re: nvidia-390xx AUR package discussion thread

#54 2021-09-06 10:35:04

Re: nvidia-390xx AUR package discussion thread

#55 2021-11-04 11:10:44

Re: nvidia-390xx AUR package discussion thread

#56 2021-11-10 08:16:09

Re: nvidia-390xx AUR package discussion thread

#57 2021-11-10 08:22:46

Re: nvidia-390xx AUR package discussion thread

#58 2021-11-10 08:30:05

Re: nvidia-390xx AUR package discussion thread

#59 2021-11-10 08:51:13

Re: nvidia-390xx AUR package discussion thread

#60 2021-11-10 08:59:06

Re: nvidia-390xx AUR package discussion thread

#61 2021-11-10 14:03:15

Re: nvidia-390xx AUR package discussion thread

#62 2021-11-10 16:07:56

Re: nvidia-390xx AUR package discussion thread

#63 2021-11-10 16:12:52

Re: nvidia-390xx AUR package discussion thread

#64 2021-11-10 17:21:22

Re: nvidia-390xx AUR package discussion thread

#65 2021-11-13 17:53:14

Re: nvidia-390xx AUR package discussion thread

#66 2021-11-13 18:55:38

Re: nvidia-390xx AUR package discussion thread

#67 2021-11-13 19:23:09

Re: nvidia-390xx AUR package discussion thread

#68 2021-11-13 20:29:17

Re: nvidia-390xx AUR package discussion thread

#69 2021-11-14 02:09:03

Re: nvidia-390xx AUR package discussion thread

#70 2021-11-14 07:48:55

Re: nvidia-390xx AUR package discussion thread

#71 2021-11-14 12:32:14

Re: nvidia-390xx AUR package discussion thread

#72 2021-11-17 18:49:37

Re: nvidia-390xx AUR package discussion thread

#73 2021-11-17 21:05:04

Re: nvidia-390xx AUR package discussion thread

#74 2021-11-18 08:44:17

Re: nvidia-390xx AUR package discussion thread

#75 2021-11-18 09:24:05

Re: nvidia-390xx AUR package discussion thread

Board footer