Tutorial - Install proprietary Nvidia driver in EasyOS

[Federico]

So, I've just finished to restore my backup (got all my settings back) and repeat the whole procedure:

- Updated the system
- Downloaded and activated kernel sources and devx SFS files
- Compiled and installed the Nvidia proprietary driver (with the --no-opengl-files option)
- Downloaded and activated the new Nvidia SFS file
- Updated the init script with the version provided above

I'm happy to confirm that everything works fine now: no problems at all! Both the desktop and the laptop boot flawlessly and I have 3D acceleration on both PCs. The desktop boots with the Nvidia proprietary driver, as expected, since the nouveau driver gets automatically blacklisted.

[BarryK]

don470,
Yes, the initrd file has to be modified, as I posted.
The problem is the change I made to the 'init' script, as in this blog post, is wrong:

https://bkhome.org/news/202311/abort-lo ... ssing.html

One line needs fixing:

Code: Select all

lspci | grep -qF 'Class 0300: 10de:'

...0300 identifies that this is a gpu, 10de identifies that it is an nvidia chip.

don570 and Federico,
There may be another problem. What changes does that NVIDIA Setup app make?
Does it change /etc/X11/xorg.conf, or create a file in /etc/X11/xorg.conf.d?
We need to check that whatever changes it makes, does not break Xorg when bootup the usb-stick on computer with intel gpu.

But even if it does make changes, it may be ok, as /etc/X11/xorg.conf and /etc/X11/xorg.conf.d are "hardware profiled" -- you will see those files/folders with numbers, for example I have xorg.conf.8086:3e91_ which is for intel gpu.

[don570]

I didn't know about the 'without-opengl-files' option
https://forums.developer.nvidia.com/t/u ... iles/33988
___________________________________________________

[BarryK]

- Downloaded and activated kernel sources and devx SFS files
- Compiled and installed the Nvidia proprietary driver (with the --no-opengl-files option)
- Downloaded and activated the new Nvidia SFS file
- Updated the init script with the version provided above

Why did you have to compile the nvidia kernel drivers?
They are already in the sfs.

[don570]

Does it change /etc/X11/xorg.conf

Whenever I look at this file it's a file written by BarryK so I don't think nvidia changes this file.
_____________________________________________________

Note that there is a file that I noticed was created during my original compiling process. I noted it at bottom of my first post.
It looks a lot like a xorg.conf file.
Location:
/usr/share/X11/xorg.conf.d/nvidia-drm-outputclass.conf

screenshot-nvidia-xorg.png (78.24 KiB) Viewed 948 times

___________________________________________________

[BarryK]

Does it change /etc/X11/xorg.conf

Whenever I look at this file it's a file written by BarryK so I don't think nvidia changes this file.
_____________________________________________________

Note that there is a file that I noticed was created during my original compiling process. I noted it at bottom of my first post.
It looks a lot like a xorg.conf file.
Location:
/usr/share/X11/xorg.conf.d/nvidia-drm-outputclass.conf

screenshot-nvidia-xorg.png
___________________________________________________

Thanks for the feedback.
That file is ok, it won't interfere when bootup the usb-stick on computer with intel gpu.

[Federico]

Sorry guys by me it's night and I've slept. Now it's 05:40 AM. I know that Barry is australian, no idea about where Don570 is from.

Using my new restored and updated system at the moment. Also made a new full backup in case of need.

Well, I had not understood that the file nvidia_5.15.138-525.147.05_amd64.sfs contains the driver itself. I asked for that but I did not get any answer.
Communication is the most important thing in any company, and even if we are not a company, we will not be very efficient without good communication imo.

So basically now I don't know if my system uses the driver which I manually compiled or the driver from the sfs file.
I can only report that Xorgwizard provides me the following options to choose from (when booting from the desktop):

- Nvidia
- nv
- modesetting
- Vesa

nouveau is missing, as expected, 'cause it gets automatically blacklisted by the sfs file.
Somebody should make another test, without manually compiling the driver and just using the sfs file and modifying the 'init' script. If it works, that would basically mean that the whole tutorial at the beginning of this thread has been made useless (although the kernel sources and devx sfs packages are very useful for making the system capable of compiling sources, generally speaking).

Regarding the Nvidia X Server Settings application, I don't know exactly where it does save its settings. I don't think it changes anything inside /etc/X11. I'll try to verify this, because I always need to manually set 36 on underscan for correcting my overscan. By the way: since I always need to repeat the setting at every boot, I think that it doesn't basically save anything: it just writes the gathered information on RAM for the current session.

Another couple of things I can report:

1) By me even audio through HDMI is working. I use an LG 55'' HD TV and I just have to choose the correct option on MCSW Soundcard Assistant.

2) A small collateral issue by me is that, when running on Nvidia proprietary driver, the system seems to loose the ability of visualizing anything on screen outside of the X server.
For example: if I type Ctrl+Alt+F2, I should get to the command prompt but I only get a dark-grey screen instead, and I have to type Ctrl+Alt+F3 in order to go back to the X server; when I power off or restart the system I cannot read anything: I just see this whole dark-grey screen and I have to trust the fact that the system is doing the right things for me.
The same does not seem to happen when using the nouveau driver, which also works, as discovered, starting from its latest iteration, but that's another story.

[BarryK]

Well, I had not understood that the file nvidia_5.15.138-525.147.05_amd64.sfs contains the driver itself. I asked for that but I did not get any answer.
Communication is the most important thing in any company, and even if we are not a company, we will not be very efficient without good communication imo.

The information about the content of the nvidia sfs was provided a few times, for example in page 1 of this thread:

viewtopic.php?p=102833#p102833

I have created a script to compile the nvidia kernel modules, X11 driver, and create an SFS with that and the GL libraries. This script will be run every time I compile the Linux kernel.

And in blog post:

https://bkhome.org/news/202311/kernel-5 ... piled.html

Linux kernel 5.15.137 compiled, as well as NVIDIA kernel modules and a NVIDIA SFS created with modules, X11 driver and GL libraries.

The name of the sfs tells you what kernel drivers are in it. nvidia_5.15.138-525.147.05_amd64.sfs is for the 5.15.138 kernel.

Anyway, a lot of information gets posted, so it is easy to miss some details.

[Federico]

Yes for sure it's also my fault not to have read everything and really everything in very depth. Sorry for that.

[bigpup]

A suggestion in the name.

Make it this way:

nvidia_525.147.05-k5.15.138_amd64.sfs

A letter k at the beginning of the specific kernel version info.

This way the numbers you first read are the driver version.
Which answers in your mind what nvidia driver?

The k tells you this is the kernel it was made for.

[Caramel]

Perhaps one solution would be, to solve the usb stick booting on different computers, for the 'initrd' to detect if there is no nvidia gpu, and if so then not load the nvidia sfs.

Maybe in the 'init' script, run the busybox lspci utility:

Code: Select all

# busybox lspci | grep 'Class 10de:'

Reference:
https://envytools.readthedocs.io/en/lat ... pciid.html

I have a nvidia GPU (Geforce 605 that needs the driver 390.157 according to https://www.nvidia.com/Download/index.aspx)
After the update to 5.6.4, the new nvidia driver has been already downloaded. It was not installed.
After the installation by SFSGet and reboot it was not loaded (the module nouveau was)

Test with lspci

Code: Select all

# busybox lspci | grep -F 'Class 10de:'

# busybox lspci | grep -F '10de:'
01:00.0 Class 0300: 10de:1048
01:00.1 Class 0403: 10de:0e08

(As usual, sorry for my English)

[Federico]

Perhaps one solution would be, to solve the usb stick booting on different computers, for the 'initrd' to detect if there is no nvidia gpu, and if so then not load the nvidia sfs.

Maybe in the 'init' script, run the busybox lspci utility:

Code: Select all

# busybox lspci | grep 'Class 10de:'

Reference:
https://envytools.readthedocs.io/en/lat ... pciid.html

I have a nvidia GPU (Geforce 605 that needs the driver 390.157 according to https://www.nvidia.com/Download/index.aspx)
After the update to 5.6.4, the new nvidia driver has been already downloaded. It was not installed.
After the installation by SFSGet and reboot it was not loaded (the module nouveau was)

Test with lspci

Code: Select all

# busybox lspci | grep -F 'Class 10de:'

# busybox lspci | grep -F '10de:'
01:00.0 Class 0300: 10de:1048
01:00.1 Class 0403: 10de:0e08

(As usual, sorry for my English)

Now you are confusing the nouveau driver with the Nvidia proprietary driver. The latest version of the nouveau driver also works but doesn't allow fine tuning, doesn't provide good performance and probably also no audio through HDMI.
The Nvidia SFS file contains the proprietary driver.
The proprietary driver may be either manually compiled (but the user looses 3D acceleration when using Easy on non-Nvidia hardware) or installed through the SFS.
Now: if one is just interested in letting Easy boot on Nvidia hardware, one can just use the integrated nouveau driver and there's no need of installing the proprietary driver, that's obvious.
But this thread was about the installation of the proprietary driver on Easy.

[Caramel]

I have a nvidia GPU (Geforce 605 that needs the driver 390.157 according to https://www.nvidia.com/Download/index.aspx)
After the update to 5.6.4, the new nvidia driver has been already downloaded. It was not installed.
After the installation by SFSGet and reboot it was not loaded (the module nouveau was)

Test with lspci

Code: Select all

# busybox lspci | grep -F 'Class 10de:'

# busybox lspci | grep -F '10de:'
01:00.0 Class 0300: 10de:1048
01:00.1 Class 0403: 10de:0e08

(As usual, sorry for my English)

I wasn't clear enough.

The test in init

Code: Select all

# lspci | grep -qF 'Class 10de:'

with busybox lspci, prevents use of the nvidia sfs on my computer although there is a nvidia card in it

EDIT :
I commented the part with the test in init and rebooted.
This time the nvidia driver was loaded :

Code: Select all

# lsmod | grep nouveau
# lsmod | grep nvidia
nvidia_drm             40960  1
nvidia_modeset       1040384  5 nvidia_drm
nvidia              15773696  153 nvidia_modeset
drm_kms_helper        184320  1 nvidia_drm
drm                   417792  4 drm_kms_helper,nvidia_drm

[Federico]

Ehm, the system should work in such a way that a red message appears during boot, saying that no Nvidia GPU has been detected and therefore the Nvidia SFS won't be loaded, when Easy is started on non-Nvidia hardware.
When using Easy on Nvidia hardware, and the Nvidia card IS detected, no red message should appear during boot and the SFS should get loaded.
Now, I don't know exactly what has been done on the init script.
The system has for sure been slightly modified in order for it to detect my particular card, which is an Asus RTX 3080.
I don't know / I'm not sure if it can also detect other RTX or all the RTX Nvidia hardware. I think this was like an experiment, it shoul be a work in progress. What card do you have? Can you post the output of lspci and busybox lspci?

[Caramel]

Ehm, the SFS works in such a way that a red message appears during boot, saying that no Nvidia GPU has been detected and therefore it won't be loaded, when Easy is started on non-Nvidia hardware.
When using Easy on Nvidia hardware, and the Nvidia card IS detected, no red message should appear during boot and the SFS gets loaded.
Now, I don't know exactly what has been done on the init script.
The system has for sure been slightly modified for detecting my particular card, which is an Asus RTX 3080.
I don't know / I'm not sure if it can also detect other RTX or non RTX Nvidia hardware. I think this was like an experiment, it's a work in progress. What card do you have? Can you post the output of lspci and busybox lspci?

It's a Geforce 605

Code: Select all

# lspci
00:00.0 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 14h Processor Root Complex
00:04.0 PCI bridge: Advanced Micro Devices, Inc. [AMD] Family 14h Processor Root Port
00:11.0 SATA controller: Advanced Micro Devices, Inc. [AMD/ATI] SB7x0/SB8x0/SB9x0 SATA Controller [AHCI mode] (rev 40)
00:12.0 USB controller: Advanced Micro Devices, Inc. [AMD/ATI] SB7x0/SB8x0/SB9x0 USB OHCI0 Controller
00:12.2 USB controller: Advanced Micro Devices, Inc. [AMD/ATI] SB7x0/SB8x0/SB9x0 USB EHCI Controller
00:13.0 USB controller: Advanced Micro Devices, Inc. [AMD/ATI] SB7x0/SB8x0/SB9x0 USB OHCI0 Controller
00:13.2 USB controller: Advanced Micro Devices, Inc. [AMD/ATI] SB7x0/SB8x0/SB9x0 USB EHCI Controller
00:14.0 SMBus: Advanced Micro Devices, Inc. [AMD/ATI] SBx00 SMBus Controller (rev 42)
00:14.1 IDE interface: Advanced Micro Devices, Inc. [AMD/ATI] SB7x0/SB8x0/SB9x0 IDE Controller (rev 40)
00:14.2 Audio device: Advanced Micro Devices, Inc. [AMD/ATI] SBx00 Azalia (Intel HDA) (rev 40)
00:14.3 ISA bridge: Advanced Micro Devices, Inc. [AMD/ATI] SB7x0/SB8x0/SB9x0 LPC host controller (rev 40)
00:14.4 PCI bridge: Advanced Micro Devices, Inc. [AMD/ATI] SBx00 PCI to PCI Bridge (rev 40)
00:14.5 USB controller: Advanced Micro Devices, Inc. [AMD/ATI] SB7x0/SB8x0/SB9x0 USB OHCI2 Controller
00:15.0 PCI bridge: Advanced Micro Devices, Inc. [AMD/ATI] SB700/SB800/SB900 PCI to PCI bridge (PCIE port 0)
00:15.2 PCI bridge: Advanced Micro Devices, Inc. [AMD/ATI] SB900 PCI to PCI bridge (PCIE port 2)
00:16.0 USB controller: Advanced Micro Devices, Inc. [AMD/ATI] SB7x0/SB8x0/SB9x0 USB OHCI0 Controller
00:16.2 USB controller: Advanced Micro Devices, Inc. [AMD/ATI] SB7x0/SB8x0/SB9x0 USB EHCI Controller
00:18.0 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 12h/14h Processor Function 0 (rev 43)
00:18.1 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 12h/14h Processor Function 1
00:18.2 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 12h/14h Processor Function 2
00:18.3 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 12h/14h Processor Function 3
00:18.4 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 12h/14h Processor Function 4
00:18.5 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 12h/14h Processor Function 6
00:18.6 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 12h/14h Processor Function 5
00:18.7 Host bridge: Advanced Micro Devices, Inc. [AMD] Family 12h/14h Processor Function 7
01:00.0 VGA compatible controller: NVIDIA Corporation GF119 [GeForce 605] (rev a1)
01:00.1 Audio device: NVIDIA Corporation GF119 HDMI Audio Controller (rev a1)
06:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 06)

Code: Select all

# busybox lspci
00:18.3 Class 0600: 1022:1703
00:13.2 Class 0c03: 1002:4396
00:18.1 Class 0600: 1022:1701
00:14.5 Class 0c03: 1002:4399
00:16.2 Class 0c03: 1002:4396
00:13.0 Class 0c03: 1002:4397
00:04.0 Class 0604: 1022:1512
00:14.3 Class 0601: 1002:439d
00:16.0 Class 0c03: 1002:4397
01:00.0 Class 0300: 10de:1048
00:14.1 Class 0101: 1002:439c
00:18.6 Class 0600: 1022:1716
00:12.2 Class 0c03: 1002:4396
00:00.0 Class 0600: 1022:1510
00:18.4 Class 0600: 1022:1704
00:15.2 Class 0604: 1002:43a2
00:12.0 Class 0c03: 1002:4397
06:00.0 Class 0200: 10ec:8168
00:18.2 Class 0600: 1022:1702
00:15.0 Class 0604: 1002:43a0
00:18.0 Class 0600: 1022:1700
00:14.4 Class 0604: 1002:4384
01:00.1 Class 0403: 10de:0e08
00:14.2 Class 0403: 1002:4383
00:11.0 Class 0106: 1002:4391
00:18.7 Class 0600: 1022:1719
00:14.0 Class 0c05: 1002:4385
00:18.5 Class 0600: 1022:1718

[Federico]

It looks like it's not supported by the driver itself:

https://www.nvidia.com/download/driverR ... 920/en-us/

We are only talking about Nvidia RTX / GTX / MX cards here. Desktop graphics cards. Yours is a particular mobile video chip, if I'm not mistaken. That should require a particular driver just for itself. That's outside of the scope of the purpose of this thread.

[BarryK]

I have a nvidia GPU (Geforce 605 that needs the driver 390.157 according to https://www.nvidia.com/Download/index.aspx)
After the update to 5.6.4, the new nvidia driver has been already downloaded. It was not installed.
After the installation by SFSGet and reboot it was not loaded (the module nouveau was)

@Caramel
Thanks for the report.

Could you clarify for me. You had the 390.* sfs loaded in easy 5.6.3, but after updating to 5.6.4 it no longer loaded.

But then you commented-out that line in the 'init' and it then did load?

At first bootup of 5.6.4, it checks if the same kernel is used in Easy 5.6.4 and if so, then will "carry over" the nvidia sfs to 5.6.4.
Well, that is what it is supposed to do.

The result of that "carry over" is you should have:

1.
/mnt/wkg/releases/easy-5.6.4/nvidia.sfs which is a symlink to the actual sfs file.

2.
Also there should be this entry in /mnt/wkg/releases/easy-5.6.4/configuration:

Code: Select all

EASY_LAYER_R01='nvidia.sfs'

The code that creates those 1. and 2. steps is at line 1860 in the 'init' script in the initrd. If that isn't working then maybe it has a typo or something.

[Federico]

@ Caramel

Oh well! It is the other SFS file Nvidia_5.15.138-390.157_amd64.sfs . Sorry I didn't know neither what that is for nor that there were already utilizers for that.

[Caramel]

@Federico , , thanks for your help

Could you clarify for me. You had the 390.* sfs loaded in easy 5.6.3, but after updating to 5.6.4 it no longer loaded.

But then you commented-out that line in the 'init' and it then did load?

At first bootup of 5.6.4, it checks if the same kernel is used in Easy 5.6.4 and if so, then will "carry over" the nvidia sfs to 5.6.4.
Well, that is what it is supposed to do.

The result of that "carry over" is you should have:

1.
/mnt/wkg/releases/easy-5.6.4/nvidia.sfs which is a symlink to the actual sfs file.

2.
Also there should be this entry in /mnt/wkg/releases/easy-5.6.4/configuration:

Code: Select all

EASY_LAYER_R01='nvidia.sfs'

The code that creates those 1. and 2. steps is at line 1860 in the 'init' script in the initrd. If that isn't working then maybe it has a typo or something.

@BarryK
I have the 1 and 2 of your message , but as i have installed the nvidia sfs by SFSGet, i don't know how it was just after the update of EasyOS

There was no problem with nvidia in EasyOS 5.6.3. The problem starts with the update of EasyOS.

In the init of Easy 5.6.4, you have added a test (https://bkhome.org/news/202311/abort-lo ... ssing.html) to prevent the loading of the nvidia sfs if no nvidia card is present in the computer.

This test fails with my card (Geforce 605). The result of busybox | grep -F 'Class 10de:' is "nothing.

I have tested the boot after commenting the lines of the test in init, and in this case the nvidia sfs is loaded.

PS : 10de: is present in the result of busybox lspci

Code: Select all

# busybox lspci | grep -F '10de:'
01:00.0 Class 0300: 10de:1048
01:00.1 Class 0403: 10de:0e08

[don570]

nvidia gt 1030 card

Code: Select all

# busybox lspci
00:1f.2 Class 0104: 8086:2822
00:1c.0 Class 0604: 8086:1c10
00:1f.0 Class 0601: 8086:1c4e
00:01.0 Class 0604: 8086:0101
00:16.0 Class 0780: 8086:1c3a
01:00.0 Class 0300: 10de:1d01
00:1b.0 Class 0403: 8086:1c20
00:19.0 Class 0200: 8086:1502
00:1e.0 Class 0604: 8086:244e
00:1f.3 Class 0c05: 8086:1c22
00:00.0 Class 0600: 8086:0100
00:1a.0 Class 0c03: 8086:1c2d
00:16.3 Class 0700: 8086:1c3d
00:1d.0 Class 0c03: 8086:1c26
01:00.1 Class 0403: 10de:0fb8
00:1c.2 Class 0604: 8086:1c14

One occurence of a card
01:00.0 Class 0300: 10de:1d01

nvidia_5.15.137-525.147.05_amd64.sfs worked with easyos 563
nvidia_5.15.138-525.147.05_amd64.sfs didn't work with easyos 564
_______________________________________

[BarryK]

I have the 1 and 2 of your message

Good, so the nvidia sfs did "carry over" to 5.6.4

The only problem is got rejected in the initrd.
Which can be fixed, as documented earlier.

I tested the kernel source sfs in easy 5.6.3.
Updated to 5.6.4, and it did carry-over, was loaded in 5.6.4
So the mechanism works.