[Solved] nova core probing fails with -524

[Nima0908]

[/b]Hello,
i need to use the nova driver for my new gentoo musl/llvm install as i have a nvidia gpu and iam running into quite and issue. After finding out what gpu i have, novacore probe instantly fails with -524 (-524 is timeout as far as iam aware). I checked if all the required firmware is installed, it indeed is. I looked through the lsmod (which returned nothing as i installed nova into the kernel not as a module), read through dmesg, which only included these 2 lines related to it:

Code: Select all

[0.559973] NovaCore 000:01:00.0: NVIDIA (Chipset: AD107, Architecture: Ada, Revision: a.1)
[0.559997] NovaCore 000:01:00.0: probe with driver NovaCore failed with error -524

I also looked at lspci which told me that the gpu was correctly recognised and also claimed by novacore. I looked at the powerstate and they were set correctly to on. I then then set loglevel=7 and drm.debug=0x1ff in cmdline and rebooted to get more verbose logs, which did absolutely nothing. I looked at the implementation of nova in the kernel code manually, but i couldnt pinpoint where it was failing. I was guessing its firmware loading, but iam not sure.

Code: Select all

dmesg | grep -i firmware

also returned nothing of use (it returned nothing at all actually). Does someone know why this happens or how i can pinpoint it? Ive never debugged drivers in the linux kernel before (only in my own, but there iam the expert so its easy there ) so i dont really know what else i can try. (Also sorry for making so many post atm. I try to do most of the troubleshooting myself, but some stuff iam just not competent enough for and i encounter a giant amount of issues atm installing a custom config llvm/musl system for the first time ). Thanks for all your answers and time

[Nima0908]

I now recompiled the kernel so that they are both modules. Both load but are not used by anything. Even with that i couldnt find out why.

[pingtoo]

I don't have any nVIDIA stuff so what about to said is pure guessing.

Do you have a directory or symbolic at {/lib,/usr/lib}/firmware/nvidia/ad107 directory or symolic?

if you missing "ad107" directory you can try to create that directory then create a symbolic link for gsp as

Code: Select all

cd /usr/lib or /lib
mkdir firmware/nvidia/ad107
ln -s ../ad102/gsp firmware/nvdia/ad107/

See if this help.

I see

Code: Select all

    35	pub(crate) const FIRMWARE_VERSION: &str = "570.144";
    36	
    37	/// Requests the GPU firmware `name` suitable for `chipset`, with version `ver`.
    38	fn request_firmware(
    39	    dev: &device::Device,
    40	    chipset: gpu::Chipset,
    41	    name: &str,
    42	    ver: &str,
    43	) -> Result<firmware::Firmware> {
    44	    let chip_name = chipset.name();
    45	
    46	    CString::try_from_fmt(fmt!("nvidia/{chip_name}/gsp/{name}-{ver}.bin"))
    47	        .and_then(|path| firmware::Firmware::request(&path, dev))
    48	}

And from

Bug#1112208: firmware-nvidia-graphics: Missing gsp symlink for nvidia ad107
Philipp Kolmann Wed, 27 Aug 2025 06:49:23 -0700

Package: firmware-nvidia-graphics
Version: 20250808-1
Severity: important

I have a NVIDIA Corporation AD107 [GeForce RTX 4060] (rev a1) in my
Debian SID box and it started failing to start GDM after installing
kernel 6.12.38 and 6.16.3.

After some debugging I found, that this newer version of the nouveau
kernel module is missing the gsp firmware blob:

kernel: nouveau NVIDIA AD107 (197000a1)
kernel: nouveau firmware: failed to load nvidia/ad107/gsp/gsp-570.144.bin (-2)

I fixed this by symlinking the ad102/gsp folder into the ad107 folder:

cd /usr/lib/firmware/nvidia/ad107/
ln -s ../ad102/gsp/

Could you do this per default? Maybe/Probably other the other ad1XX
directories need this folder with newer nouveau versions as well?

thanks
Philipp

[Nima0908]

Thank you for your answer. Sadly, this is not the sollution. ad107 is just a symlink to ad102, and it contains all files (i checked in both /usr/lib and /lib).

[Nima0908]

I asked one of the developers, and they told me the reason for this is that it aint ready to use yet.