Nouveau module not loading on Kernel 6.14+

Ask Fedora
, , , , ,
1

Running Fedora 41 Atomic, with a Nvidia 3080Ti GPU, trying to get the default Nouveau driver to load with Secure Boot Enabled.

Everything works fine with Kernel 6.13. However with both 6.14 and 6.15 the nouveau module is not loading, based on output of lsmod. I tested 6.14 with both Fedora 41 and 42, 6.15 with just F41, the issue is the same regardless of the Fedora version. To get module loading I override my rpm-ostree deployment with latest 6.13 kernel, leaving everything else in the image the same, and all is good. Seems likely to be an issue in the kernel packages based on all of that.

Looking at After upgrade to F42, I cannot load VirtualBox drivers signed modules in the kernel (SecureBoot) - #13 by cbr84 I see that both the xz compression and signing hash algo changed in 6.14.

It seems like the xz params are correct, but that the signing hash algo is using sha256, not sha512. However, I cherry-picked some modules that are loaded and checked, and both (wireguard and nf_tables) are also using sha256 signing hash algo, so not sure if this is a red herring:

$ xz -l -vv /lib/modules/6.15.3-100.fc41.x86_64/kernel/drivers/gpu/drm/nouveau/nouveau.ko.xz
/lib/modules/6.15.3-100.fc41.x86_64/kernel/drivers/gpu/drm/nouveau/nouveau.ko.xz (1/1)
  Streams:           1
  Blocks:            3
  Compressed size:   1,090.8 KiB (1,117,016 B)
  Uncompressed size: 8,272.4 KiB (8,470,950 B)
  Ratio:             0.132
  Check:             CRC32
  Stream Padding:    0 B
  Streams:
    Stream    Blocks      CompOffset    UncompOffset        CompSize      UncompSize  Ratio  Check      Padding
         1         3               0               0       1,117,016       8,470,950  0.132  CRC32            0
  Blocks:
    Stream     Block      CompOffset    UncompOffset       TotalSize      UncompSize  Ratio  Check      CheckVal  Header  Flags        CompSize    MemUsage  Filters
         1         1              12               0         567,804       3,145,728  0.181  CRC32      d5362ce7      20  cu            567,779       2 MiB  --lzma2=dict=1MiB
         1         2         567,816       3,145,728         238,660       3,145,728  0.076  CRC32      e4301238      20  cu            238,636       2 MiB  --lzma2=dict=1MiB
         1         3         806,476       6,291,456         310,500       2,179,494  0.142  CRC32      507cd3ce      20  cu            310,473       2 MiB  --lzma2=dict=1MiB
  Memory needed:     2 MiB
  Sizes in headers:  Yes
  Minimum XZ Utils version: 5.0.0

$ modinfo nouveau
vermagic:       6.15.3-100.fc41.x86_64 SMP preempt mod_unload 
sig_id:         PKCS#7
signer:         Fedora kernel signing key
sig_key:        5C:E5:EF:7A:8F:F3:3C:C3:C1:43:89:D5:55:E5:C5:7F:71:7B:E2:53
sig_hashalgo:   sha256

Trying to load the module manually fails:

# modprobe -vv nouveau
modprobe: INFO: libkmod/libkmod.c:391 kmod_set_log_fn() custom logging function 0x555fd1db5220 registered
insmod /lib/modules/6.15.3-100.fc41.x86_64/kernel/drivers/gpu/drm/nouveau/nouveau.ko.xz 
modprobe: INFO: libkmod/libkmod-module.c:973 kmod_module_insert_module() Failed to insert module '/lib/modules/6.15.3-100.fc41.x86_64/kernel/drivers/gpu/drm/nouveau/nouveau.ko.xz': Operation not permitted
modprobe: ERROR: could not insert 'nouveau': Operation not permitted
modprobe: INFO: libkmod/libkmod.c:358 kmod_unref() context 0x555fe35104c0 released

The nouveau module is in the initramfs image:

$ lsinitrd /boot/ostree/fedora-3997c8e689d5605d8799f99c040bc3fd55c84fe8da3246740825f2eb12d0379d/initramfs-6.15.3-100.fc41.x86_64.img | grep -i nouveau
drwxr-xr-x   2 root     root            0 Dec 31  1969 usr/lib/modules/6.15.3-100.fc41.x86_64/kernel/drivers/gpu/drm/nouveau
-rw-r--r--   1 root     root      1117016 Dec 31  1969 usr/lib/modules/6.15.3-100.fc41.x86_64/kernel/drivers/gpu/drm/nouveau/nouveau.ko.xz

Looking at dmesg warn level and above, and journalctl -k with warning level and above nothing is jumping out that is different compared to the 6.13 kernel.

Not sure if this is related, but another regression with kernel 6.14+ is failing to unmount certain mounts on shutdown. systemd-shutdown reports:

[FAILED] Failed unmounting etc.mount - /etc.
[FAILED] Failed unmounting usr.mount.
[ 185.018471] systemd-shutdown[1]: Unable to finalize remaining DM devices, ignoring.
2

Issue might be related to nouveau being replaced by new rust based nova_core drivers in newer kernels.

3

But I don’t think nova core is usable yet.

Someone else said they had to black list nova core to allow nouveau to be loaded.

4

it was enabled by default on my Fedora KDE system after upgrading to 6.15 kernel today. I too had to blacklist it and nouveau both since i use Negativo17’s nvidia drivers.

5

If I’m right and it’s not ready for use then maybe the kernel should not be building it yet.
You could report as a bug against the Fedora kernel.

6

Nova was first included in 6.15 though, not 6.14. So it wouldn’t explain why this issue started for 6.14 for me.

Possible though that blacklisting Nova + whatever changes are in 6.15 will fix it though, will try when I’m at my computer again.

7

Yep, confirmed that blacklisting nova doesn’t fix the issue. Nouveau module is still not loading.

9

is nvidia-gpu-firmware installed on your system?

# uname -r
6.15.3-200.fc42.x86_64

# rpm -qa nvidia-gpu-firmware
nvidia-gpu-firmware-20250613-1.fc42.noarch

# lsinitrd  |grep -c firmware/nvidia
494

this is on F42 Workstation booted wi/o blacklist for nouveau AND nova_core

# cat /proc/cmdline 
BOOT_IMAGE=(hd1,gpt5)/vmlinuz-6.15.3-200.fc42.x86_64 root=UUID=85c7ead2-fd7d-4168-8094-2d666ca20d35 ro rootflags=subvol=root quiet

# lsmod | grep -E 'nouveau|nova|nvidia'
i2c_nvidia_gpu         12288  0
nova_core              20480  0
nouveau              3874816  37
drm_ttm_helper         16384  3 xe,nouveau
gpu_sched              65536  2 xe,nouveau
drm_gpuvm              45056  2 xe,nouveau
drm_exec               12288  3 drm_gpuvm,xe,nouveau
i2c_algo_bit           20480  3 xe,i915,nouveau
ttm                   135168  4 drm_ttm_helper,xe,i915,nouveau
mxm_wmi                12288  1 nouveau
drm_display_helper    311296  3 xe,i915,nouveau
video                  81920  3 xe,i915,nouveau
wmi                    32768  4 video,wmi_bmof,mxm_wmi,nouveau

blacklist is only required for the nvidia driver.

# modinfo nouveau
vermagic:       6.15.3-200.fc42.x86_64 SMP preempt mod_unload 
sig_id:         PKCS#7
signer:         Fedora kernel signing key
sig_key:        30:1D:E5:12:25:A8:63:D7:68:FB:6F:8A:A8:33:DF:AE:9B:56:A1:B7
sig_hashalgo:   sha256
10

It should work with blacklisting nova-core:

sudo rpm-ostree kargs --append=rd.driver.blacklist=nova-core --append=modprobe.blacklist=nova-core
11

Yes nvidia-gpu-firmware is installed, it’s part of the rpm-ostree image which I haven’t modified.

And yes I already blacklisted nova-core, but that didn’t resolve anything which is expected since this issue began with kernel 6.14, before nova-core module was added.

1 Like