Upgrade from Debian10 to Debian11: issue with the grafic card Radeon [1002:6611]
 

Hello,
since I have upgraded to Debian11 yesterday, few errors appear in dmesg and the grafic card dont send data straight into my screen via HDMI switch (I have to on/off the switch now).

If anybody has an idea what to do, let me know (this is a small issue; by clicking the HDMI switch on/off I see Debian11 on the screen).


lspci -nn
...
...

dmesg | grep -i "radeon"

Not much to be done I would think, the new driver does not like the switch connection, you need to have it connect again. The only way I could see for a resolution to your problem is get in touch with the developers and see if they want to resolve the new behaviour by debugging it with you.

Thanks. Not much a big problem. However, I see a difference: it comes straight after i915

Old DMESG: (Debian10)
New DMESG: (Debian11)
Could it be the i915 onboard grafic and the radeon would disturb together?
Why a non primary device (i915) was tried to get a setup?
i915 grafic perhaps should be blacklisted. I will further search.

The blacklist idea is worth a try, it is not being used so no sense in it even loading.

Taking i915 out would be an issue if the radeon grafic would scratch?
https://wiki.debian.org/KernelModuleBlacklisting
I would not like to mess up my system with an initramfs command where I could not go back.
Further finding; It looks like the i915 in Debian11 has a more extended setup; see below.
Comments/remarks are welcome.

Debian10 kernel
grep -i i915 /boot/config-4.19.0-17-amd64
Debian11 kernel
grep -i i915 /boot/config-5.10.0-8-amd64

As well it should that is many version of the kernel later, there is bound to be some progress on the driver. It is up to you if you want to try but there is not much to lose on a one time boot with the module blacklisted. You do not use the onboard card so it will still use the add in card. All you have to do to reverse it is eliminate the blacklist file do the dependmod command again and update the initramfs once more. Reboot to have everything back to where it was before you did the black listing.

I had to blacklist and make a fake install. i915 gone. Issue still there. Old errors still there. One new errors coming (no effect seen in the sound).
I still have to click on my HDMI switch (the card still to wake up due to this; else the screen goes into a sleep mode). A bit ennoying.
Look like not an i915 and radeon interaction but now a radeon issue on Debian11. I will try to identify again the dmesg output when I switch the HDMI after a while.

Update.

PC boot this morning after a night when off: this was a cold boot. Screen is coming after a while (a bit too long in my opinion; whatever thats ok).

a) error below are gone

[ 2.894117] [drm:radeon_dp_link_train [radeon]] *ERROR* clock recovery tried 5 times
[ 2.894270] [drm:radeon_dp_link_train [radeon]] *ERROR* clock recovery failed

b) yesterday I made following (effect positive on a cold boot?)

apt-get update
apt-get dist-upgrade ; one lib was updated (cannot remember which one)
pip3 install amdgpu-fan ; see https://wiki.debian.org/AtiHowTo
apt-get install radeontop ; see https://wiki.debian.org/AtiHowTo

After a reboot, the issue is back again.

So, there is a difference between cold boot and warm boot. I had such an issue few time ago (that was sound; not graphic), I will see if a similar approach will solve https://www.linuxquestions.org/quest...ve-4175534057/

This PC has a Win10 double boot. No issues at all with Win10.

Any advice/help is welcome.

Adding as boot parameter made it https://wiki.archlinux.org/title/ATI

cat /proc/cmdline
Thread closed.

And the Arch Linux wiki strikes again, they have some good pages on that thing, I have used it a few times as well. Good you hear you got it going and nice to see the solution posted.

Yes. Arch Wiki is a source of good things. My feeling is, this is not the first time I am having a solution from there. With your post I came with the radeon.pcm=0 which made it. But at a third boot, nomore. And then I was further searching for parameters radeon.xxx=1/0. Due to cold boot / warm boot observed behaviour it made sense an instability in the hardware had to compensated. A pci boot parameter made sense for me. I tried. It works 5 times. Not longer. Situation is still unstable.

Closed. Done.

lspci -nn
Radeon HD 8570 is "Sea Islands" https://en.wikipedia.org/wiki/Radeon#Sea_Islands
Card advice https://wiki.debian.org/AtiHowTo#gcn1011
By implementing it as boot parameter (see below).
cat /proc/cmdline
System is ok. Why it was not necessary for Debian10? I dont know.

You now boot a newer kernel with Debian 11 unless you had the backports enabled in Debian 10 and went out of your way to install a newer kernel, that would be the difference. I had to do this to have my ZFS work in the 10 version when it stopped working on the upgrade from the 9. Luckily I had waited enough time to do the upgrade for the bugs to have been filed so I got to find the solution quickly there. Then once I went with the 11 I forgot the backports was in a file in the sources.list.d directory and it broke again with the old files it installed from there.

Issue came again today despite it worked yesterday. Let keep this thread closed. I will come back here if I have found a solution with a stable situation for few weeks.

I am still battling with my card after the upgrade. 2 screens can be attached but I did not find precisely the sequence to make this happing continously.Any advice/idea is welcome.

Following error happens now:

Not sure what it is.

dmesg | grep -i "error"
Kernel command line: BOOT_IMAGE=/boot/vmlinuz-5.10.0-8-amd64 root=UUID=300f2560-ac9d-4eb9-8d18-c155909766c7 ro quiet 8250.nr_uarts=6 radeon.pcie_gen2=0 radeon.dpm=0 radeon.cik_support=0 amdgpu.cik_support=1

I will observe how stable it is the next days. So far ok. Re-reading the Arch Wiki was a good thing.