cannot get a working kernel build

[mark4]

well sort of can but... weird stuff

Just built a 6.18.7 kernel from gentoo sources and at boot it gets as far as loading initramfs and I get no further console output. If however I proceed to type in my / partitions decryption key my drive led starts to flash as if loading... still no console output at all. When the flashing stops and I type in my login I again see some flashy blinky and startx launches X.

However, nvidia-drivers does not seem to be working properly because nvidia-settings has almost nothing in it, there is tearing when I move windows suggesting I am software rendering and glxgears gives me a whopping 100 170 FPS.

my kernel config which was configured by following https://wiki.gentoo.org/wiki/User:Pieti ... ent_Kernel but I probably shafu'd something in there

http://dpaste.com/G435D6JV8

[pietinger]

my kernel config which was configured by following https://wiki.gentoo.org/wiki/User:Pieti ... ent_Kernel but I probably shafu'd something in there

... not quite ...

Code: Select all

# CONFIG_DRM_I915 is not set
# CONFIG_INTEL_IOMMU_DEFAULT_ON is not set

... also ... NVIDIA ... ... with every new nvidia driver version there are new kernel settings necessary ... sometimes something must be enabled ... sometimes the same must be disabled ... try to disable this

Code: Select all

CONFIG_SYSFB_SIMPLEFB=y

it it off by default ... and can make big problems ... maybe your nvdia driver does not need it (I am also unsure if the description in our NVIDIA Wiki article is correct for this parameter ... I dont know ... it its NVIDIA )

(all other options for framebuffer driver and console are correct)

[mark4]

my kernel config which was configured by following https://wiki.gentoo.org/wiki/User:Pieti ... ent_Kernel but I probably shafu'd something in there

... not quite ...

Code: Select all

# CONFIG_DRM_I915 is not set
# CONFIG_INTEL_IOMMU_DEFAULT_ON is not set

... also ... NVIDIA ... :evil: ... with every new nvidia driver version there are new kernel settings necessary ... sometimes something must be enabled ... sometimes the same must be disabled ... try to disable this

Code: Select all

CONFIG_SYSFB_SIMPLEFB=y

it it off by default ... and can make big problems ... maybe your nvdia driver does not need it (I am also unsure if the description in our NVIDIA Wiki article is correct for this parameter ... I dont know ... it its NVIDIA :evil: )

(all other options for framebuffer driver and console are correct)

I cant remember what I did but i got output working on the fb during boot, im not 100% sure i need the intel stuff though because my laptop bios allows me to select between integrated only, discrete only or hybrid and I have it set to discrete only. compiling the intel stuff in might still be a good idea if later I decide I want to switch to using that only... I change my mind a lot :)

As it stands right now my startx takes about 10 seconds before I get anything on display and when my user attempts to run nvidia-settings it tells me that the nvidia drivers are not loaded. However, if I su to root i CAN launch nvidia-drivers and once root has done so so can my user. very weird.

< edit >

I stated the above incorrectly, user can run nvidia-settings its just that theres nothing in there at all, no gpu0 etc. That all becomes active once root runs nvidia-settings.

< /edit >

[pietinger]

I cant remember what I did but i got output working on the fb during boot, [...]

[...] im not 100% sure i need the intel stuff though because my laptop bios allows me to select between integrated only, discrete only or hybrid and I have it set to discrete only. [...]

The reason I mentioned it was: Some nvidia-drivers (in the past; today?) needed some helper-options in the kernel ... but ... these options are not selectable ... you can get it only when selecting i915 (or others like amdgpu). So, in the past there was the "trick" to enable another drm to get the needed helper-options for nvidia. Please dont ask me about the current state ... I am sorry, I cannot help with nvidia further. (you see that our LiveCD uses Nouveau ... )

[logrusx]

Does gentoo-kernel-bin work? If it does, try gentoo-kernel with savedconfig use enabled and gradually strip parts of the configuration and don't touch sections you don't understand. Have backups of the config on every step. After you reduce enough (by your judgement) of the kernel, try building it yourself. If you manage to produce working kernel, now you can drop gentoo-kernel and shave off configuration manually. Still, keep backups of the config on every step.

Best Regards,
Georgi

[mark4]

Does gentoo-kernel-bin work? If it does, try gentoo-kernel with savedconfig use enabled and gradually strip parts of the configuration and don't touch sections you don't understand. Have backups of the config on every step. After you reduce enough (by your judgement) of the kernel, try building it yourself. If you manage to produce working kernel, now you can drop gentoo-kernel and shave off configuration manually. Still, keep backups of the config on every step.

Best Regards,
Georgi

I do not believe the gentoo kernel bin will work because my / partition is LUKS encrypted but I am not using LVM at all.

Things are very weird right now, I cannot tell if I have a serious hardware problem or if the drivers are just not working for me. I do have video but if i play minecraft i get 2 frames per second and nvidia-settings is almost completely blank until i run it as root at least once. After running it as root the user can also run it but the video settings tab shows absolutely no information what so ever and the GPU-0 tab shows GPU utilization locked to 0%.

This morning when booting the machine booted straight into windows so I booted the system rescue, unlocked / which is on the same drive as windows, chrooted in, mounted /boot and cd'd to it and ran grub-install --efi-directory=EFI /dev/nvme1n1p1 and I got my grub menu back on reboot. I have seen grub become uninstalled at power down before so this was an easy fix but very worrying that it is happening.

I do have a question though my two nvme drives are nvme0n1 and nvme1n1 and they flip flop which is which so I have to do lsblk to see where drive 0 is. WHY???? whiskey tango lol!

[Hu]

Device discovery order is non-deterministic. You should not rely on any particular drive being a particular ID. Instead, use drive attributes, partition attributes, or filesystem attributes to identify the target of interest.

[mark4]

[ 19.880444] nvidia-modeset: WARNING: GPU:0: Unable to read EDID for display device DP-4
[ 19.894519] nvidia-modeset: WARNING: GPU:0: Unable to read EDID for display device DP-4
[ 19.901327] [drm] Initialized nvidia-drm 0.0.0 for 0000:01:00.0 on minor 1
[ 19.903771] nvidia 0000:01:00.0: [drm] Cannot find any crtc or sizes