Bug 25216 - dkms-nvidia-current will not start on some older cards.
Summary: dkms-nvidia-current will not start on some older cards.
Status: RESOLVED FIXED
Alias: None
Product: Mageia
Classification: Unclassified
Component: RPM Packages (show other bugs)
Version: 6
Hardware: All Linux
Priority: Normal major
Target Milestone: ---
Assignee: QA Team
QA Contact:
URL:
Whiteboard: MGA6-64-OK
Keywords: advisory, validated_update
Depends on:
Blocks:
 
Reported: 2019-08-02 21:52 CEST by Zombie Ryushu
Modified: 2019-08-10 02:14 CEST (History)
3 users (show)

See Also:
Source RPM: dkms-nvidia-current
CVE:
Status comment:


Attachments

Description Zombie Ryushu 2019-08-02 21:52:36 CEST
dkms-nvidia-current will not start on some older cards. One of my systems has an Nvidia 610 that needs the Nvidia 390 branch of drivers. Mageia forcibly updates to the 410 branch which causes X.org to loop. Eventually it gives up and switches to Noevau

Installing the 390 branch prohibits updating the Kernel from 4.15 to 4.18, the DKMS Compile fails. This could be solved if the DKMS Module made a distinction between 410 and 390.

My suggested resolution for this problem is for Mageia to package all existing legacy branches of dkms-nvidia the way Rosa does.
Comment 1 Thomas Backlund 2019-08-03 00:05:18 CEST
Sorry, but nope.

We wont package every nvidia legacy driver.
Its useless extra work for no real gain.

Officially supported kernel in mga6 is 4.14 which current drivers in mga6 support.

The kernel 4.18 is in backports mostly to help out with some newer amd gpus.

The nvidia-current R410 you only get if you have nonfree backports enabled, and it was only provided for some newer gpus, but it also drops some older gpus so its not a simple upgrade...


having said that, I guess we could provide an update to the nvidia-current R390 branch in nonfree/updates* that would should work with newer kernels...
 
But also... you do realize Mga7 is out since ~1 month, and Mga6 hits EOL in ~2 months, so you'd be better off upgrading to Mga7 where you would be shifted to the nvidia390 driver for the lifetime of Mga7

CC: (none) => tmb
Assignee: bugsquad => kernel

Comment 2 Zombie Ryushu 2019-08-03 00:56:29 CEST
As a stop gap, doing the R390 upgrade would keep my Nvidia 610 working in Mageia 7, I am going to upgrade systems to Mageia 7 soon, but I need to ensure my hardware will remain supported.
Comment 3 Thomas Backlund 2019-08-03 11:01:48 CEST
According to https://nvidia.custhelp.com/app/answers/detail/a_id/3142/~/support-timeframes-for-unix-legacy-gpu-releases

The Linux 390.* legacy driver series is the last to support GF1xx ("Fermi") GPUs.  Support for new Linux kernels and X servers, as well as fixes for critical bugs, will be included in 390.* legacy releases through the end of 2022.

and a driver branch does not drop hw suppoort, so it should be good for  mga7 and mga8
Comment 4 Thomas Backlund 2019-08-03 11:53:18 CEST
I've pushed  nvidia-current-390.129-1.mga6 to Mga6 nonfree updates testing
Comment 5 Thomas Backlund 2019-08-03 11:58:18 CEST
Avisory to follow...

So the rpms to test:


SRPMS:
nvidia-current-390.129-1.mga6.nonfree.src.rpm


i586:
dkms-nvidia-current-390.129-1.mga6.nonfree.i586.rpm
nvidia-current-cuda-opencl-390.129-1.mga6.nonfree.i586.rpm
nvidia-current-devel-390.129-1.mga6.nonfree.i586.rpm
nvidia-current-doc-html-390.129-1.mga6.nonfree.i586.rpm
x11-driver-video-nvidia-current-390.129-1.mga6.nonfree.i586.rpm


x86_64:
dkms-nvidia-current-390.129-1.mga6.nonfree.x86_64.rpm
nvidia-current-cuda-opencl-390.129-1.mga6.nonfree.x86_64.rpm
nvidia-current-devel-390.129-1.mga6.nonfree.x86_64.rpm
nvidia-current-doc-html-390.129-1.mga6.nonfree.x86_64.rpm
x11-driver-video-nvidia-current-390.129-1.mga6.nonfree.x86_64.rpm

Assignee: kernel => qa-bugs

Comment 6 Morgan Leijström 2019-08-04 20:59:16 CEST
mga6-64, 390.129-1.mga6, Plasma

Updated, had to go, came back much later, screen saver had kicked in and I and was greeted with a black screen and white text that said unlocking had failed and i should run "loginctl unlock-session c2" from another terminal.

I Ctrl-Alt-F1 to that terminal and saw it repeatedly output a screen full of NVIDIA messages amongst them "API mismatch".  It is not very user friendly to switch to tty and input that command being interrupted by scrolling messages...

Is there some way to tell the user that reboot is needed for continued reliable operation (or was there, an i forgot i had seen it?)

CC: (none) => fri

Comment 7 Morgan Leijström 2019-08-05 12:52:31 CEST
Continuation of comment 6: An additional quirk is that at least on my Plasma system, as usual after such upgrades the shutdown and reboot entries from many do not work.  We have to i.e open Konsole and issue "reboot".

----

Rebooted, it works OK:
video+audio in Firefox, Thunderbird, LibreOffice6, ...
CUDA and OpenCL recognized by BOINC, but not tested in use.

This system also updates all installed to testing.

$ uname -a
Linux svarten 4.14.131-desktop-1.mga6 #1 SMP Thu Jun 27 11:19:36 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux

Hardware: i7-2600K, Nvidia GTX760 (GK104) using proprietary driver GeForce 420 and later.
Thomas Backlund 2019-08-09 23:20:52 CEST

Whiteboard: (none) => MGA6-64-OK
Keywords: (none) => advisory, validated_update
CC: (none) => sysadmin-bugs

Comment 8 Morgan Leijström 2019-08-09 23:42:30 CEST
For Comment 6, I entered Bug 25251
Comment 9 Mageia Robot 2019-08-10 02:14:10 CEST
An update for this issue has been pushed to the Mageia Updates repository.

https://advisories.mageia.org/MGAA-2019-0089.html

Status: NEW => RESOLVED
Resolution: (none) => FIXED


Note You need to log in before you can comment on or make changes to this bug.