1. Please describe the problem: After upgrading to kernel 6.15.3-200.fc42, my MacBook6,1 with NVIDIA GeForce 9400M (using the nouveau driver) fails to boot into the graphical interface. The screen freezes with colored vertical lines and graphical corruption immediately after the GRUB selection, and the system becomes unresponsive. Booting into the previous kernel (6.14.11-300.fc42) works fine. I suspect a regression in nouveau or related DRM changes that affect older NVIDIA hardware. 2. What is the Version-Release number of the kernel: 6.15.3-200.fc42 3. Did it work previously in Fedora? If so, what kernel version did the issue first appear? Yes, the issue did not occur in: 6.14.11-300.fc42 6.14.0-63.fc42 The problem first appeared in 6.15.3-200.fc42. 4. Can you reproduce this issue? If so, please provide the steps to reproduce the issue below: Yes, it is 100% reproducible. Steps: Boot Fedora 42 on a MacBook6,1 with NVIDIA GeForce 9400M (G96M) GPU. Let the system load kernel 6.15.3-200.fc42. The display becomes corrupted with colored blocks/lines. System freezes before graphical login (Hyprland or any DE). Rebooting into 6.14.11 works normally. 5. Does this problem occur with the latest Rawhide kernel? I have not tested the Rawhide kernel yet, but I can do it if necessary and provide logs. 6. Are you running any modules that are not shipped directly with Fedora's kernel? No, I am using only the default nouveau driver shipped with Fedora. No NVIDIA proprietary drivers or DKMS modules are installed. 7. Please attach the kernel logs. I'm attaching the output of: journalctl --no-hostname -k -b -1 > dmesg.txt (The -b -1 refers to the previous boot attempt where the issue occurred.) System information: OS: Fedora Linux 42 (KDE Plasma Desktop Edition) x86_64 Host: MacBook6,1 (1.0) Kernel: Linux 6.14.11-300.fc42.x86_64 Display: 1280x800 @ 60Hz (Built-in LCD) WM: Hyprland 0.45.2 (Wayland) CPU: Intel Core 2 Duo P7550 @ 2.26GHz (2 cores) GPU: NVIDIA GeForce 9400M [Discrete] — using nouveau RAM: 7.5 GiB Swap: 7.5 GiB Filesystem: btrfs Workaround: Use previous kernel (6.14.11). Add exclude=kernel* in /etc/dnf/dnf.conf to block updates temporarily. Remove 6.15.3 manually to restore system. Reproducible: Always
Does it work if you boot with modprobe.blacklist=nova_core on the kernel command line?
Created attachment 2095160 [details] dmesg.txt from failed boot with kernel 6.15.3-200.fc42 (nouveau issue)
I tested booting kernel 6.15.3-200.fc42 with and without the modprobe.blacklist parameter. 🔴 In both cases: The system hangs during early systemd initialization (on the Fedora splash screen). No TTYs are available (Ctrl+Alt+F2 does nothing). I have to do a forced shutdown using the power button to recover. 🟢 Booting kernel 6.14.11-300.fc42 works flawlessly, including: Graphical environment (Hyprland, KDE Plasma + nouveau). TTYs (Ctrl+Alt+F2, etc.). No boot issues or freezes. This confirms a serious regression in kernel 6.15.3, possibly related to nouveau or low-level system initialization, affecting MacBook6,1 with NVIDIA GeForce 9400M.
Update: I first tried booting kernel 6.15.3-200.fc42 with the kernel parameter modprobe.blacklist=nouveau to isolate the issue. This removed the graphical corruption, but the system still froze on the Fedora splash screen (spinning circle), and TTYs were inaccessible. A hard shutdown was required. Afterwards, I tried booting with modprobe.blacklist=nova_core, and surprisingly, the system booted completely fine: No graphical issues. Reached the graphical session (Hyprland and KDE Plasma on Wayland). TTYs worked properly. This behavior is consistent and reproducible. It suggests the presence of this kernel parameter — even with a non-existent module — alters the boot timing or module initialization enough to avoid the issue. This could indicate a race condition or low-level init regression in kernel 6.15.3 affecting nouveau or DRM. Kernel 6.14.11-300.fc42 continues to work flawlessly without any kernel parameters.
It is interesting that nova_core blacklist fixes the issue as I saw no mention of it in dmesg. The module actually is present on the system, it is new with 6.15, but it is not particularly useful yet. Eventually it will replace both nouveau and the proprietary nvidia module. It is enabled because it is a driver in rust and we want to make sure that it is building as expected, and allow testing with it as new functionality is introduced. Outside of that, there are a couple of nouveau issues which I know are being addressed in 6.15.4 which should release tomorrow.
I have a VERY old machine with GeForce 6150SE nForce 430 integrated video, which works normally in 6.14.11, drops to emergency mode in 6.15.3, but boots and runs normally with modprobe.blacklist=nova_core. Should I file a separate bug?
(In reply to Andre Robatino from comment #6) > I have a VERY old machine with GeForce 6150SE nForce 430 integrated video, > which works normally in 6.14.11, drops to emergency mode in 6.15.3, but > boots and runs normally with modprobe.blacklist=nova_core. Should I file a > separate bug? No need to file a separate bug. I need to think on this a bit. We basically want nouveau taking precedence over nova_core until we don't. Meaning nova should eventually replace both the nouveau and the proprietary nvidia drivers. We also want nova_core building even before it is useful to ensure that things still build correctly (as one of the few rust drivers). Of course nova shouldn't ever really support the older nvidia GPUs at all AFAIK, so I am not sure why it is claiming those PCI Ids.
Hey. I get nouveau failure with 6.15.3 and .4 as well. dmesg: nouveau 0000:01:00.0: NVIDIA AD103 (193000a1) nouveau 0000:01:00.0: gsp ctor failed: -2 nouveau 0000:01:00.0: probe with driver nouveau failed with error -2 That's with or without "rd.driver.blacklist=nova_core modprobe.blacklist=nova_core" as suggested in various discussions. 6.14.11 works normally.
I have same issue as Jens Falsmar Oechsler. RXT 4070 TI, 6.14 worked fine with nouveau. 6.15 - nouveau fails to load. Adding nova_core to backlist - does not help. ``` [ 3.458193] nouveau 0000:01:00.0: probe with driver nouveau failed with error -2 ```
With kernel-6.15.5-200.fc42.x86_64, my machine that dropped to emergency mode with 6.15.3 and 6.15.4 now instead just hangs at "Starting systemd-vconsole.setup.service - Virtual Console Setup...". With "modprobe.blacklist=nova_core", works normally as before.
(In reply to deishelon+redhat from comment #9) > I have same issue as Jens Falsmar Oechsler. > RXT 4070 TI, 6.14 worked fine with nouveau. 6.15 - nouveau fails to load. > Adding nova_core to backlist - does not help. > > ``` > [ 3.458193] nouveau 0000:01:00.0: probe with driver nouveau failed with > error -2 > ``` This issue with newer Nvidia hardware is resolved with new firmware (nvidia-gpu-firmware-20250708).
The changelog for kernel-6.15.10-200.fc42 says "- Disable NOVA_CORE (Justin M. Forbes)" and I confirmed that with this kernel, I don't need to blacklist nova_core on my affected machine. (But I'm keeping "modprobe.blacklist=nova_core" as a default boot argument for now until things stabilize.)
What is the status of this? As of 6.17.4 I can still boot my old machine without blacklisting nova_core.
We are keeping nova_core disabled until it is actually useful for the hardware it is meant to cover.