I noticed that but figured it had to do with my messing around. Yes, go ahead and remove it from the repo. I don’t want my fooling around to cause others issues.
I tested more today and the issue is networkmanager not shutting down. If I disable
networkmanager it will reboot just fine.
Can you post a kernel config where I can get to it. What kernel version is it?
Unfortunately it is 5.8.15-301
https://drive.google.com/file/d/1fJ9ZrNSkni6bVyNqC3B6Oor3C2UDhLQX/view?usp=sharing
Oh. So I can just use the exact instructions you gave me before? Excellent. Will try that tonight.
linux-rpi4-5.10.23-1 seems to break fkms with my 4K monitor and lightdm. I get the dreaded blinking cursor in the upper left that xorg is not running properly. Changing my config.txt to use kms gives the same result. So I guess EDID is broken again. I rolled back to 5.10-22 to get it working again.
I have been messing with the kernel configs and got some more kernel configs to show up. I am re-compiling 5.10.23 for you to test. Seems like you were using a minimal install to test uefi with aren’t you.
Yes, I am using the minimal install image, minus the /boot of course. I replaced it with what the uefi firmware provided.
@0n0w1c Try this kernel on your uefi minimal test image.
https://drive.google.com/file/d/1ISFmx7tMmnAOSFJ652oz4dv-d65FDMKx/view?usp=sharing
I will test the new kernel in a couple of minutes.
I updated to the 5.10.23-1 kernel again. This time I added back the hdmi_group and hdmi_mode to my config.txt, and that works with fkms.
The test 5.10.23 kernel build boots via UEFI shell!
1 Like
I am not positive about this, but I think ACPI support will be required in the kernel config. The UEFI firmware did not have SD or WiFi support prior to 5.12. This was due to the SD and wifi drivers lacking ACPI bindings, as I recall. So I assume it will be a required kernel config to get them to work.
Here is a link to the closed ticket on the SD. Maybe you can make better sense of it than I can. 
Edit: This is probably a better link, the actual patch. In the comments it says:
With this change, Linux is capable of utilizing the SD card slot, and the
Wi-Fi when booted with UEFI+ACPI on the RPi4.
And the code is wrapped inside of a #ifdef.
#ifdef CONFIG_ACPI
Correct. That is what pulled in a ton of modules in the last test 5.10 kernel above.
I now have a gpt partitioned SSD loaded with plasma, fbturbo, and arm-unstable on a btrfs rootfs that boots via UEFI to a grub menu that can select a Manajro ARM kernel with a dracut initramfs. And it seems to be a functional! 
The current raspberrypi-bootloader package seems to work and can safely overwrite the loader provided by the UEFI firmware. So I have pacman set to ignore only the kernel and headers. Unfortunately, so far no working v3d or sound, but I have not exhausted my efforts.
1 Like
Could you please provide a tuto to let everyone have a chance to test it ?
I will try to post something, hopefully later today.
Same old reboot issue on the latest 5.12
Something else that seems to work with UEFI support added:
Just a proof of concept, but it shows promise… the basics work, on both x86_64 and aarch64 kvm hosts. And that is a Manjaro RPi4 image that I used for the kvm guest disk image.
1 Like
I was afraid of that. I am thinking it is likely something in wpa_supplicant that needs adjusting.
I am thinking it is likely something in wpa_supplicant that needs adjusting.
Me too. Looking at the wpa_supplicant package it is really old.
I removed the wpa_supplicant bin from /usr/bin so the service would not load and it rebooted just fine.
1 Like