Frogging-Family / nvidia-all

Nvidia driver latest to 396 series AIO installer
733 stars 64 forks source link

Black Screen #120

Closed DarkXero-dev closed 1 year ago

DarkXero-dev commented 1 year ago

Hey...

After rebuilding 515-dkms drivers at r214 and installing getting black screen with AMD Kernel.. I dunno why ! On mainline Arch Kernel no issues booting to OS ..

Tk-Glitch commented 1 year ago

What's "AMD Kernel"? Anyway if it works with Arch kernel it's not an issue with nvidia-all. Nvidia modules can't be compiled against any kernel, and if interfaces used by them were modified in a way that's no handled, it will fail.

DarkXero-dev commented 1 year ago

It used to work, was working fine on r211 when I updated to r214 it broke.. Also I use dkms .. And AMD kernel is a Ryzen optimized.. Something happened between r212 n r214 that caused it to break ?

Tk-Glitch commented 1 year ago

First, nvidia-all aren't custom drivers. You get what Nvidia provides. Secondly, 212 introduced 515.57, 213 introduced 515.49.06 (vulkan dev) and 214 introduced 515.49.10 (vulkan dev). There was exactly zero change to anything we handle between 212 and 214. If you're still using the same driver revision and the same choice regarding open-source kernel modules, the package is effectively exactly the same.

DarkXero-dev commented 1 year ago

When I get prompted about beta I always build saying No.. So regular is built.. For some reason caused black screen .. From what you are saying I cannot use custom Kernels anymore with latest ?

Tk-Glitch commented 1 year ago

I've been using 515.57 and 515.49.10 fine on linux-tkg 5.18, so it's something else that broke it for you.

DarkXero-dev commented 1 year ago

I use 515.57 not 515.49.10 I did nothing special.. All I did was update to r214 nothing more

Tk-Glitch commented 1 year ago

Also, there are many nvidia-all + custom kernel users out there, and you're the first to report such an issue. Nobody pinged me on Discord either.

Tk-Glitch commented 1 year ago

I use 515.57 not 515.49.10 I did nothing special.. All I did was update to r214 nothing more

r214 didn't change anything for 515.57, so either your kernel is broken or something turned bad in the chain somewhere. But no change in nvidia-all for sure.

DarkXero-dev commented 1 year ago

Like I said all I did was build 515.57 r214 and update nothing more now I get black screen.. And talked to AMD Kernel maintainer he said he did nothing special.. Could it be missing patches ?

Tk-Glitch commented 1 year ago

If we changed anything in there I'd gladly track the bug out, but in this case there's simply no change :/

DarkXero-dev commented 1 year ago

Strange.. Nature of Arch I guess.. Will wait for next AMD Kernel update n use mainline one in meantime.. No biggie, keep this open will reply with any changes ..

Thanks ;)

Tk-Glitch commented 1 year ago

Also, if you don't believe me and wanted to go back to r212, you can git checkout aaf1bc0dbc91a4b29f9f096a3936363f9930f499 in your nvidia-all clone.

Tk-Glitch commented 1 year ago

Could you link that "AMD Kernel" source if possible? The fact that it works with Arch's (and linux-tkg custom kernel, as tested by many) tends to point to an issue with that specific kernel.

DarkXero-dev commented 1 year ago

Nahh it's ok will wait for Kernel update, if nothing works I will have to roll back.. Problem is I host them on my repo for my users.. In case anyone else reports issue will let you know...

AMD Kernel source (AUR) https://aur.archlinux.org/packages/linux-amd

Tk-Glitch commented 1 year ago

Thanks. Indeed it doesn't seem to do much. Graysky's patch and.. That's about it? If it's otherwise a plain copy of Arch's kernel there should be no difference in behavior (building with -march=whatever doesn't do much). That's definitely weird.

DarkXero-dev commented 1 year ago

My feeling exactly.. I am not a dev at all just a regular user who just knows how to build a custom Distro.. So I do not understand code xD

ptr1337 commented 1 year ago

I think just the kernel needs a rebuild. GCC has been updated the last 3 days to 12.0.3 which maybe results into that your dkms is failing. Just reinstall the kernel and check if your dkms that output any error - if yes rebuild the kernel

DarkXero-dev commented 1 year ago

I think just the kernel needs a rebuild. GCC has been updated the last 3 days to 12.0.3 which maybe results into that your dkms is failing. Just reinstall the kernel and check if your dkms that output any error - if yes rebuild the kernel

I thought about that I re-installed Kernel nothing.. No errors and still black screen

Tk-Glitch commented 1 year ago

Since it enforces znver3 are you using a Zen3 CPU?

DarkXero-dev commented 1 year ago

Ryzen 5 3600x so yeah....

Tk-Glitch commented 1 year ago

That's not zen3, that's zen2. For desktop CPUs, Zen3 are 5000 series. With the retbleed mitigations in place with 5.18.14, it's not unlikely that the kernel is now more sensitive to zen3 optimizations on your zen2 CPU, breaking things. What you could try to validate that theory is to change that line in your custom kernel PKGBUILD: https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=linux-amd#n81 so it's using -march=znver2 -mtune=znver2 instead of the current -march=znver3 -mtune=znver3. Nonetheless znver3 is totally incorrect for your CPU.

DarkXero-dev commented 1 year ago

Ok so it's Zen 2 damn so it's now enforced. Was working before... Flawless... I guess will have to rebuild using this https://aur.archlinux.org/packages/linux-amd-znver2

Good catch thanks...

Tk-Glitch commented 1 year ago

Considering that custom kernel's changes are seemingly only related to those µarch optimizations, that's the most likely culprit. I'll be crossing my fingers for you 😄

DarkXero-dev commented 1 year ago

Will let you know. No power right now...

DarkXero-dev commented 1 year ago

Just tried znver2 for Zen2 same issue.. I am now compiling tkg Kernel with zen2 patches to see if that works.. If it fails I dunno what to do except stick with mainline Arch Kernel xD

Update : tkg Kernel worked !!!! Weird that AMD one for my CPU didn't xD

TKG

LethalManBoob commented 1 year ago

latest driver gives me black screen. had to chroot in an old version.

Tk-Glitch commented 1 year ago

Booted just fine on my end with a 980ti on Z370. Anyway this issue isn't related and I'll close it as the root cause was found :frog: