[SOLVED] Black screen on AMD APU on fresh install

Durobot · June 13, 2023, 6:01pm

Hi everyone!
I’m new to Endeavour, but have some experience with Arch, however nothing similar to this -

Installed Endeavour on a HP Elitedesk 705 G5 DM, with Ryzen 5 3400GE APU (no dedicated video card), but get black screen after “Reached target Graphical Interface” 95% of the time. Ctrl+Alt+Fkeys doesn’t work. The machine is OK, passes all hardware tests, Win 10 worked without problem, but I nuked it to install Endeavour.

Hardware information: http://ix.io/4y9S
Boot log: http://ix.io/4y9T

Enabled sshd that one time the system did actually load KDE Plasma for me, so now I can access the machine over ssh.
If I add nomodeset kernel parameter, the boot log stays on the screen (last line being ), but I can use Ctrl+Alt+F2 (F3, etc.) to log into the system.

I’m not an expert, but these lines in my boot log don’t look good to my eye:

Jun 13 20:06:23 hp705 kernel: [drm] Unsupported Connector type:21!
Jun 13 20:06:23 hp705 kernel: [drm] Unsupported Connector type:21!
Jun 13 20:06:23 hp705 kernel: [drm] Unsupported Connector type:21!
Jun 13 20:06:23 hp705 kernel: [drm] Unsupported Connector type:21!

Jun 13 20:06:23 hp705 kernel: kfd kfd: amdgpu: Failed to resume IOMMU for device 1002:15d8
Jun 13 20:06:23 hp705 kernel: kfd kfd: amdgpu: device 1002:15d8 NOT added due to errors

However, I have no idea what to make of this.

xircon · June 13, 2023, 6:23pm

This is the bit that looks odd to me:

Display: server: X.org v: 1.21.1.8 driver: X: loaded: modesetting unloaded: amdgpu gpu: amdgpu
    tty: 151x39

Have you installed:

12 extra/xf86-video-amdgpu 23.0.0-1 (69.4 KiB 166.1 KiB) [xorg-drivers] (Installed)

I am new to AMD gpus, so cannot be much help, but:

https://wiki.archlinux.org/title/AMDGPU

Might help.

Durobot · June 14, 2023, 4:55am

Yes, I do have xf86-video-amdgpu installed:

$ pacman -Q | grep xf86-video-amdgpu
xf86-video-amdgpu 23.0.0-1

List of installed packages (pacman -Q): http://ix.io/4ybC

I was able to boot into KDE Plasma once more, by pure luck - after a few failed attempts, I logged onto the box via SSH did sudo reboot, and presto. Here’s the boot log: http://ix.io/4ybB

But that was just a stroke of luck, which I can’t reproduce. Here’s another boot to black screen afterwards: http://ix.io/4ybF

Guess I’ll try making sense of the diff between these two boot logs…

Update: I have generated a comparison report for good (KDE Plasma) and bad (black screen) boot logs. Good is on the left, bad is on the right. Here’s the file:
good-bad-boot-logs.odt (85.1 KB)
Doesn’t look great as ODT, but the forum won’t let me upload an HTML file.

So, it looks like the problems starts here -
In the good boot log I have:

 hp705 kernel: amdgpu 0000:0b:00.0: amdgpu: Fetched VBIOS from ROM BAR
 . . .
 hp705 kernel: [drm] RAM width 64bits DDR4
 . . .
 hp705 kernel: fbcon: amdgpudrmfb (fb0) is primary device
 hp705 kernel: Console: switching to colour frame buffer device 240x67
 hp705 kernel: amdgpu 0000:0b:00.0: [drm] fb0: amdgpudrmfb frame buffer device
 . . .
 hp705 kernel: warning: `kded5' uses wireless extensions which will stop working for Wi-Fi 7 hardware; use nl80211

and in the bad boot log:

 hp705 kernel: amdgpu 0000:0b:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
 ... (several lines that are in the good boot log as well, but they come before this part) ...
 hp705 kernel: amdgpu 0000:0b:00.0: amdgpu: Fetched VBIOS from platform
 . . .
 hp705 kernel: [drm] RAM width 128bits DDR4
 . . .
 hp705 kernel: [drm] Unsupported Connector type:21!
 hp705 kernel: [drm] Unsupported Connector type:21!
 hp705 kernel: [drm] Unsupported Connector type:21!
 hp705 kernel: [drm] Unsupported Connector type:21!

Looks like the first sign of trouble is this message:

amdgpu 0000:0b:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff

If only I knew what causes it…

xircon · June 14, 2023, 7:20am

Have you got:

cat /etc/X11/xorg.conf.d/20-amdgpu.conf
Section "OutputClass"
     Identifier "AMD"
     MatchDriver "amdgpu"
     Driver "amdgpu"
EndSection

Also mkinitcpio or dracut? grub or systemd boot?

Durobot · June 14, 2023, 7:32am

I have installed EndeavourOS using default settings pretty much - just changed the keyboard layout and added root password, so it uses systemd-boot. DE is KDE Plasma.

Regarding /etc/X11/xorg.conf.d/20-amdgpu.conf, I believe it is there, as a part of xf86-video-amdgpu, if memory serves me right. I have not modified it, so it must be the same or very close to what you’ve posted. I can’t be 100% certain before I look at mine.
Will check it when I get home in about 8 hours.

Also mkinitcpio or dracut?

Um… probably mkinitcpio, as I went with defaults pretty much… Must check to be sure.

xircon · June 14, 2023, 7:33am

We are in different time zones obviously I am about to go to work (UK).

Durobot · June 14, 2023, 7:39am

I’m in GMT+3, so not far off.

I’ve been looking around, and it seems like this amdgpu <something>: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff thing is not new at all: https://bugzilla.kernel.org/show_bug.cgi?id=207767.

So, perhaps I could muck around in BIOS/UEFI settings, and/or add either amdgpu.dc=0 or amdgpu.dc=1 kenel flags? Will try and report back.

xircon · June 14, 2023, 7:54am

Yes, the driver is just not loading and defaulting to modesetting driver. Off to work.

mihalycsaba · June 14, 2023, 10:23am

Did you turn on legacy boot or did something else with your UEFI or hardware? That error looks like something’s up with your UEFI. Try resetting it to default settings, if it doesn’t work, then reflash the latest firmware.

Durobot · June 14, 2023, 10:28am

In fact, I did turn on Enable legacy boot, disable secure boot, IIRC, as before that I could not boot from USB thumb drives.
I’ll try changing that setting when I get home.

Thank you for the tip!

mihalycsaba · June 14, 2023, 10:30am

Try a different program for making the install media. I’ve been using rufus with default settings for years now, barely any problem.

mihalycsaba · June 14, 2023, 10:32am

Don’t turn on legacy mode if it’s not necessary, but if your firmware lets it, it’s okay to turn off secure boot.

Durobot · June 14, 2023, 10:34am

IIRC, legacy mode and secure boot are combined in one, and you can only select from a list of combinations. I just wanted to turn off secure boot, but it came as a package with legacy mode.

Something like this (not my picture, my BIOS could be different if similar):
c05869166

mihalycsaba · June 14, 2023, 10:37am

Yeah I saw a firmware where you couldn’t turn off secure boot on it’s own. Just leave it enabled then, also restore your settings to default, hope it helps…

ricklinux · June 14, 2023, 1:40pm

 amdgpu 0000:0b:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff

This message is related to the pci bus and is harmless.

hp705 kernel: [drm] Unsupported Connector type:21!

Edit: Is this a laptop?

Durobot · June 14, 2023, 2:55pm

Not a laptop, but an HP EliteDesk 705 G5 DM (Desktop Mini).
It’s a small PC, some call them 1L (one liter) PCs. Looks like this:

Regarding this connector thing… I wonder if it could be related to the way these machines output video signal. You see, it has 2 display port connectors on the back (DP++ variety), but they also have a connector for a small “Flex IO” board. These boxes practically always come with one of these boards installed, which could be one of VGA out, HDMI out, USB, Ethernet, etc.

Mine came with a VGA board, but I wanted HDMI, and so instead of DP to HDMI adapter I bought an HP HDMI Port Flex IO board. For the record, these boards are not the same for Intel and AMD-based HP EliteDesk machines, so mine is specific to 705 model, and its part number is 3TK75AT.
It looks like this:
41q8G70B5CL.AC_SX466

You don’t think this is the connector this message actually refers to?

Also, EndeavourOS live / installation USB always boots to XFCE perfectly well, video signal coming out of this Flex IO board nicely.

Durobot · June 14, 2023, 5:21pm

So, I’ve changed my Configure Legacy Support and Secure Boot BIOS / UEFI setting from “Legacy Support Enable and Secure Boot Disable” to “Legacy Support Disable and Secure Boot Disable”, and it helped!

For the record, here are screenshots of my BIOS before (i.e. when I had the problem):

All available options:

The option I selected, which solved the problem:

Big thanks to everybody in this thread!
Guys, you’re great!

ricklinux · June 14, 2023, 10:27pm

So is it working now with changing the Bios settings?

Durobot · June 15, 2023, 4:40am

Yes, the change from “Legacy Support Enable and Secure Boot Disable” to “Legacy Support Disable and Secure Boot Disable” has solved the problem. I was able to boot into KDE Plasma (not sure if X11 or Wayland, TBH) 3 times in a row.

I always thought “Legacy Support” affects the way USB boot media is handled - GPT/MBR, or something like that. I would never imagine it affects iGPU / APU. Go figure.

ricklinux · June 15, 2023, 3:53pm

Wayland is not the default. You need plasma-wayland-session in order to log in with wayland. I also would not have thought “Legacy Support Enable and Secure Boot Disable” would have caused this.