Can not compile Nvidia driver in BookwormPup64 10.0.6

[bigpup]

To get a dmesg I could read and find stuff.

Had to use this:

Code: Select all

dmesg ls | less

with this you can scroll thru the lines of messages.

Yes I found a line that says the driver is loaded.
It is close to the end of dmesg output.

It is this:
loading NVIDIA UNIX X86_64 kernel module 525.147.05

This seems to be what was done by doing:

Code: Select all

modprobe nvidia-current

In a earlier part of dmesg output it has this:
nvidia loading out of tree module taints kernel

Still no help

Still running xorgwizard and entering nvidia as the driver to use has no affect to get it to boot past the point of needing to run xorgwizard.
No driver selections you choose will get it to boot and start X server.
Not even vesa driver.

[dimkr]

@bigpup Can you share /var/log/Xorg.0.log after failed start of X, when the selected driver is nvidia and after modprobe nvidia-current?

[Clarity]

@dimkr you have several Vanilla (Debian) WoofCE distro versions which are reasonably similar to BKWP64. Could you take one and do the steps you indicate that 'work' using your distro.

We can test and try mapping those steps to compare steps here with the distro you have where the steps works?

OR, as a guide, could you give a utility that does the steps you feel would generate the outcome we're seeking?

We could use that as a current and future base for adding nvidia to/in a Debian based WoofCE Pup like BKWP64.

If you think it helpful and When time permits, of course.

Edit: Just back from doctor visit and have notice I have made a mistake in member name selection. Corrected the member name.

[radky]

Yes I found a line that says the driver is loaded.
It is close to the end of dmesg output.

It is this:
loading NVIDIA UNIX X86_64 kernel module 525.147.05

@bigpup

Just a few thoughts:

When upstream Debian Bookworm 12.5 was initially released with kernel 6.1.76 there were many reports of failed integration with the nvidia 525.147.05-4 driver available in primary Debian repositories at that time. This was subsequently fixed when the nividia 525.147.05-7 driver was added to the bookworm-updates repository.

For me, a quick download today (21 Mar 2034) confirms APT provides the updated 525.147.05-7 nvidia package (rather than the older problematic 525.147.05-4 package), but I don't know if that was the case when you installed nvidia-driver a few days ago.

For confirmation, you could check log files in the /var/log/apt directory

[mikewalsh]

@Clarity :-

Um.....who on EARTH is "Dimaspaf14"? (I see there's a member by that name, yes.....and all his posts are in the VanillaDPup forum. What makes you think it's our Dima [dimkr]?)

Just curious, that's all...

Mike.

[dimkr]

failed integration with the nvidia 525.147.05-4 driver available in primary Debian repositories at that time

I guess this refers to:

* Apply pfn_valid patch from gentoo to fix kernel module build for
Linux 6.1.76, 6.6.15, 6.7.3, 6.8. (Closes: #1063363, #1062932)

(https://metadata.ftp-master.debian.org/ ... _changelog)

From what I see, this is irrelevant.

The driver consists of two parts. First is the nvidia-current.ko kernel driver, and apparently it's installed correctly (so the problem is not build failure of the driver, which was a problem before this Debian package update) and it loads fine according to modprobe nvidia-current.

Second part is the nvidia X.Org driver, which apparently fails. The unique thing about Puppy is xorgwizard (therefore, this is the "thing" where I expect the problem to be once we isolate it): other distros don't it, don't have something equivalent to it or simply don't have xorg.conf and simply rely on the default auto-detection (or just dropped X.Org completely and went Wayland-only). /var/log/Xorg.0.log after failed start of X with the nvidia driver specified in xorg.conf will tell why the X.Org driver doesn't work.

[bigpup]

Yes I found a line that says the driver is loaded.
It is close to the end of dmesg output.

It is this:
loading NVIDIA UNIX X86_64 kernel module 525.147.05

@bigpup

Just a few thoughts:

When upstream Debian Bookworm 12.5 was initially released with kernel 6.1.76 there were many reports of failed integration with the nvidia 525.147.05-4 driver available in primary Debian repositories at that time. This was subsequently fixed when the nividia 525.147.05-7 driver was added to the bookworm-updates repository.

For me, a quick download today (21 Mar 2034) confirms APT provides the updated 525.147.05-7 nvidia package (rather than the older problematic 525.147.05-4 package), but I don't know if that was the case when you installed nvidia-driver a few days ago.

For confirmation, you could check log files in the /var/log/apt directory

This what it shows in my /var/log/apt/

Install: update-glx:amd64 (1.2.2, automatic), nvidia-alternative:amd64 (525.147.05-7~deb12u1, automatic), glx-alternative-nvidia:amd64 (1.2.2, automatic), nvidia-egl-icd:amd64 (525.147.05-7~deb12u1, automatic), glx-diversions:amd64 (1.2.2, automatic), libnvidia-eglcore:amd64 (525.147.05-7~deb12u1, automatic), libnvidia-glcore:amd64 (525.147.05-7~deb12u1, automatic), libglx-nvidia0:amd64 (525.147.05-7~deb12u1, automatic), nvidia-kernel-support:amd64 (525.147.05-7~deb12u1, automatic), nvidia-driver-bin:amd64 (525.147.05-7~deb12u1, automatic), libnvidia-egl-wayland1:amd64 (1:1.1.10-1, automatic), libnvidia-ml1:amd64 (525.147.05-7~deb12u1, automatic), nvidia-driver-libs:amd64 (525.147.05-7~deb12u1, automatic), nvidia-driver:amd64 (525.147.05-7~deb12u1), nvidia-support:amd64 (20220217+3~deb12u1, automatic), nvidia-kernel-common:amd64 (20220217+3~deb12u1, automatic), libgl1-nvidia-glvnd-glx:amd64 (525.147.05-7~deb12u1, automatic), nvidia-kernel-dkms:amd64 (525.147.05-7~deb12u1, automatic), glx-alternative-mesa:amd64 (1.2.2, automatic), nvidia-egl-common:amd64 (525.147.05-7~deb12u1, automatic), nvidia-legacy-check:amd64 (525.147.05-7~deb12u1, automatic), nvidia-vdpau-driver:amd64 (525.147.05-7~deb12u1, automatic), libegl-nvidia0:amd64 (525.147.05-7~deb12u1, automatic), firmware-nvidia-gsp:amd64 (525.147.05-7~deb12u1, automatic), xserver-xorg-video-nvidia:amd64 (525.147.05-7~deb12u1, automatic), nvidia-modprobe:amd64 (535.54.03-1~deb12u1, automatic), nvidia-installer-cleanup:amd64 (20220217+3~deb12u1, automatic)

[bigpup]

@bigpup Can you share /var/log/Xorg.0.log after failed start of X, when the selected driver is nvidia and after modprobe nvidia-current?

How exactly would I get this if I cannot get booted past running xorgwizard?

We may be able to get this to finally work.

But I think all the script files that get installed are the big issue making it not work.

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

But it would be much better to be able to install by using the Nvidia driver run package from Nvidia!!!

One big reason is you can select to use different nvidia drivers.

The big question that needs to be found a fix for is why could the nvidia driver run package compile in earlier versions of BookwormPup64 10.0.3 and 10.0.4 and not compile in this one 10.0.6?????????????

I guess I could just change the kernel to the one in BookwormPup64 10.0.4 and use the pet package that was made form the compiled driver using the nvidia driver run package.
That run package compiled in 10.0.4 and also was able to make a good working driver pet package using Getnvidia Puppy program.

[dimkr]

@bigpup Can you share /var/log/Xorg.0.log after failed start of X, when the selected driver is nvidia and after modprobe nvidia-current?

How exactly would I get this if I cannot get booted past running xorgwizard?

Can you boot with pfix=ram and grab the log file from the save?

[dimkr]

I think I found the problem, it's a bug in /usr/bin/xwin. When you select the nvidia driver, it clears your selection before X is started, so you're stuck in a weird situation where the kernel driver is active but the matching X.Org driver isn't.

This line (252?):

Code: Select all

			if ! [ "`find $DRVRSPATH -type f -name "*${CURRENT_DRIVER}*"`" ] ; then

Should have been:

Code: Select all

			if ! [ "`find -L $DRVRSPATH -type f -name "*${CURRENT_DRIVER}*"`" ] ; then

EDIT: I borrowed a laptop with an nvidia GPU, and it looks like X loads the driver correctly after apt install nvidia-driver libnvidia-cfg1 nvidia-xconfig && nvidia-xconfig (to completely bypass xorgwizard and get rid of the xorg.conf generated by it)

[bigpup]

@dimkr

You got it!

That method works for getting the driver by using apt!

I hope you are going to make this edit to xwin and make it a change in next version of BookwormPup64.
Maybe offer it as a pet package of a new version of xwin until it gets updated.

One thing I had to additionally do is install the nvidia X server settings program so could have control of the drivers settings.

Code: Select all

apt install nvidia-settings

The nvidia driver is controlled by settings made with this program.
when the driver is being used all control of it is provided by a settings file.
/root/.nvidia-settings-rc

I made a topic on exactly how to do this driver install:
viewtopic.php?t=11038
.
.

[bigpup]

Still need to figure out why the Nvidia run package will not compile in this version of BookwormPup64?

I am sure this will be needed to be able to install a newer driver than the one in the Debian repositories.

[dimkr]

Still need to figure out why the Nvidia run package will not compile in this version of BookwormPup64?

It's sensitive to kernel version, including minor version, just like aufs and any other third party driver. The driver in Debian 12.x should work with the kernel in Debian 12.x, so apt is more likely to work compared to upstream driver.

Regarding the xwin bug - my fix is merged (https://github.com/puppylinux-woof-CE/woof-CE/pull/4258).

[radky]

Still need to figure out why the Nvidia run package will not compile in this version of BookwormPup64?

I am sure this will be needed to be able to install a newer driver than the one in the Debian repositories.

@bigpup

The upstream Debian nvidia 525.147.05 source code is now patched (pfn_valid patch) to compile successfully with kernel 6.1.76. This fix applies to nvidia source code (patched by Debian maintainers) available to BW64 via the APT package manager.

Presumably, at some point, upstream Nvidia will also patch their more recent video drivers (perhaps not the legacy versions) and make them available for download. Until the updates are available, the source code downloaded directly from the Nvidia website may not compile successfully with kernel >= 6.1.76).

[bigpup]

Thanks to all that helped figure out how to get the Nvidia driver installed by using apt.

Here is a how to on installing using apt:
viewtopic.php?t=11038

Good catch by dimkr finding that error in the code of xwin file.

I know me and other users with Nvidia hardware thank you all!!

[Gobbi]

Last week , I also tried using getnvidia in BookwormPup64 10.0.6 , kernel 6.1.76 without success.
I also tried to use apt but it kept saying that nouveau was still in use so .... no driver .
Then I tried to boot without X , with prefix=nox , having wired internet and then , I updated apt on command line , after that
using apt I installed nvidia-driver and nvidia settings only . It installed the 525.147.05-7 .
I didn't know anything about " libnvidia-cfg1 nvidia-xconfig && nvidia-xconfig " . Thank's for sharing @dimkr & @bigpup

But , I had enter xorgwizard , chose manually nvidia as driver , chose my monitor's resolution and 24bit colours
and in the end xwin , and I got a working desktop , driver and Nvidia settings software working . After RestartX , several times ,
the desktop came back with the Nvidia driver still working .
By the way , my card is RTX 4060 .
In cases like this I usually do a remaster to keep the changes .
I tried three times to remaster but had no success in keeping a functional working desktop with the nvidia driver in the remasterd sfs .
I don't know if this was because of the overlays file system or not . I need to learn more about it .
Thank everybody for sharing .

Edit on 25-03-2024 : Today , with a working Nvidia driver installed like I mentioned above , I found a workaround for keeping the changes .
I don't usually use a savefile , but now I tried and managed to save to a folder . Later , after reboot , I used nicOs-Utility-Suite and made a sfs out of the save-folder ,
renaming it ydrv .