Comments

59 Comments
karma

I was wrong. The amdgpu suspend fix 3a9626c816db901def438dc2513622e281186d39 is included.

karma

3a9626c816db901def438dc2513622e281186d39 was included in kernel-6.7.5-200 but accidentally dropped from this build because of miscommunication where that part of the amdgpu suspend fix was not applied to stable. This will result in suspend crashes while our 6.7.5 did not crash.

Correction: this didn't fix the ath11k suspend deadlock from 6.7.3. It seems to happen less often. It isn't worse though so this should be published.

karma

https://bugzilla.redhat.com/show_bug.cgi?id=2262577 kernel-6.7.4 fixed this ath11k_pci crash on suspend.

https://bugzilla.redhat.com/show_bug.cgi?id=2262577 This kernel broke suspend on AMD Ryzen 7 PRO 6850U with Radeon Graphics

karma

wine-9.0 stable is in good shape upstream. wine-staging contains experimental stuff. It breaks wayland mouse in some applications. Do you have a specific reason for shipping this experimental patch? Please stick to the stable mainline release?

karma

working great on intel 12th gen mobile

karma

seems good on Thinkpad P1 Gen 5 with 12th Gen Intel(R) Core(TM) i9-12900H

karma

works

karma

seems fine

kernel-6.5.5 has the same regression as kernel-6.5.[0-5] where amdgpu crashes hard with multiple generations of AMD APU's. I'm investigating further.

I was unable to reproduce the amdgpu crash in Firefox. It crashes consistently with Chrome or Chromium, the latter is available in Fedora.

Note: There is something different about Live USB where it doesn't crash. Ryzen 7 5850U with Fedora 39 Beta Workstation Live does not crash in chromium despite it having the problematic kernel-6.5.2. Unclear why it is different during Live runtime. My guess would be power management being inhibited or something.

There are linux-firmware updates in updates-testing, including for AMD. Are those relevant?

amd-gpu-firmware-20230919-1.fc38.noarch then regenerated the initramfs makes no difference.

Tested with both kernel-6.5.4 and kernel-6.5.5rc1.

https://gitlab.freedesktop.org/drm/amd/-/issues/2830 amdgpu crash regression confirmed across 3+ generations of Ryzen APU's. Please help by bringing this to the attention of AMD.

https://gitlab.freedesktop.org/drm/amd/-/issues/2830 6.5.x including 6.5.4 have this serious amdgpu regression. You can crash it while zooming Google Maps in chromium.

karma

working great on AMD Ryzen 7 PRO 5850U with Radeon Graphics

works

https://bugzilla.redhat.com/show_bug.cgi?id=2236516#c11

As discussed here the problem turned out to be chrome/chromium generated code caches causing the browser to malfunction after certain mesa/llvm updates.

https://bugzilla.redhat.com/show_bug.cgi?id=2193335 People will have unexpected surprises from their browser but this isn't llvm's fault. chromium needs to learn to invalidate its generated code caches.

Resubmitting llvm for updates-testing.

karma

Confirmed this is a duplicate of Bug #2193335. Chrome/Chromium with GPU acceleration can be broken after mesa upgrade because its cached objects become incompatible.

https://discussion.fedoraproject.org/t/chromium-based-browsers-display-garbled-web-pages-after-mesa-is-updated/83438 rm ~/.config/google-chrome/GrShaderCache -rf rm ~/.config/google-chrome/ShaderCache -rf rm ~/.config/google-chrome/Default/GPUCache -rf Deleting these cache files fixes the browser.

So this isn't mesa's fault. This mesa update should go to stable.