Repository navigation
9070XT: Worked in VT (1080P@60Hz) but not in Plasma #6
Description
I've attached 2 dmesgs.
This one is booting with my adapter and then switching to the HDMI port:
And that one is booting straight with the HDMI port:
I don't know if it's my TV or my adapter, but one of them is peculiar because anytime I try such thing, I have trouble to get them working together again... But regular HDMI-HDMI works fine on regular kernels now unlike with your tree. The adapter worked fine with your tree.
Thank you!
Reacted by Daniel
Activity
mkopec commentedon Feb 19, 2026
The Adding stream 00000000856cde73 to context failed with err 31 errors should be fixed as of 714f865 . Not sure about not working in Plasma, though, Might be worth a try
John-Gee commentedon Feb 19, 2026
I think the difference between Plasma and vt was the resolution and frequency, but with latest I could not get vt nor Plasma working though I only tried with booting on the adapter and then switching, I'll try the other way in a bit:
after.txt
No better straight:
straight.txt
I doubt it matters, but just in case, I use booster to generate the initrd not mkinitcpio.
John-Gee commentedon Feb 20, 2026
I saw you had more patches since so I tried straight on HDMI again but still nothing.
Now dmesg is a lot more verbose so I hope the fact that it cutoff the first 30s is ok with you:
straight.txt
mkopec commentedon Feb 20, 2026
hmm, the continued retraining requests are a bit suspicious, and it may be hard to fix. But it does finish training eventually, so it should work...
John-Gee commentedon Feb 21, 2026
The various requests may be because I kept trying to switch vt and gui seeing if something would show?
If it finishes training and you think it should work, is there something to try on the TV to help maybe?
Thank you!
John-Gee commentedon Feb 22, 2026
I just realized your latest changes are not in the staging branch but in the regular one so I tried again with the 6.19 branch, this time just by booting with the adapter and then switching as I saw someone in another bug saying that was the only way but still no go.
I also simplified my kernel cmd line incase some of the amdgpu options were interfering but no better.
This log is complete and with a stack trace so hopefully more useful:
after.txt
John-Gee commentedon Feb 22, 2026
Out of curiosity I tried again, turning my TV on and off, and disabling the forced resolution for the VT (it helps with adapter, without that the TV often stays black when switching to VT). Somehow this ended up restarting the machine so maybe there is something there too?
restart.txt
John-Gee commentedon Feb 23, 2026
Related to #17 : This is also on a TCL TV, mine is a Roku one, not sure how much this matters.
perf cs-etm: Fix decoding for sparse CPU maps
perf lock contention: fix segfault in `lock contention -b/--use-bpf`
mkopec commentedon Feb 24, 2026
Would you mind retrying on the latest code? I pushed some changes to the FRL rate picking and retraining logic
John-Gee commentedon Feb 25, 2026
I've just tried and it didn't seem different.
But looking at other issues it got me thinking of switching my HDMI settings in my TCL TV.
I switched from Standard (which means latest, and I had ALLM and VRR on) to Auto and now I get a display!
Alas it's 4K@60Hz with no VRR or HDR available in Plasma's monitor settings, so same as without your patches except without them Standard settings works fine here I need Auto which likely defaulted to Compatibility which means HDMI 1.4 instead of 2.1.
after.txt
Just in case I picked the wrong branch:
❯ uname -a
Linux 6.19.3-00101-g55b86007e517 #11
hdmi_frl_stable git branch as master is now on 7.0rc
Out of curiosity I tried switching back the HDMI setting on the TV to Standard but without turning on ALLM and VRR, that gave no picture again.
I'm attaching the log, it's the same boot but it seems the earlier lines were cut so it might be useful to have both files.
after.txt
John-Gee commentedon Feb 25, 2026
Would it be useful to you if I tried on a computer monitor instead of a TCL TV? I don't have one, but I should be able to borrow one for a bit if it can help. Just wondering if there'd be any data you could use that way, sort of separating the quirks from the gpu from the tv's ones.
mkopec commentedon Feb 25, 2026
No need, all 9070XTs should work the same. Mine does work, so it's more of a quirk of the TV.
Can you try the latest commit? I made an adjustment that should allow training to pass
John-Gee commentedon Feb 25, 2026
Got you. In this case, running Windows with the TV? I'm not sure how I would do that though passthrough won't work with a single GPU and I don't have room in my case for 2.
So latest didn't help either, and it ended up restarting the machine again:
restart.txt
I appreciate all your help!
Would it be better to join the Matrix channel? If so is there a better time for you?
edit: So I don't know for sure that your patches are what cause the restart, I stick with the same patched kernel even when using my adapter and I don't have such restart but you never know, maybe it's something in the HDMI path that I don't use with my DP adapter that causes that.
John-Gee commentedon Feb 26, 2026
In case it helps, edid: https://gist.github.com/John-Gee/c6ed996b6e629c272400699cab8de776
spec: https://www.displayspecifications.com/en/model/ae592e2e
Merge branch 'selftests-bpf-fixes-for-userspace-asan'
selftests/bpf: Fix a memory leak in xdp_flowtable test