https://bugs.kde.org/show_bug.cgi?id=527526
[email protected] changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |[email protected] --- Comment #5 from [email protected] --- Same issue here on much older hardware and the 580xx legacy driver. Some extra data points: Setup: - Laptop MEDION Erazer X7841 (Clevo P670RE1M), CachyOS, Plasma 6.8.1, Frameworks 6.31.0, Qt 6.12.0, Wayland - Intel HD Graphics 530 (Skylake, i915) drives the internal eDP-1 display; KWin composites on it (OpenGL ES, Mesa 26.2.4) - NVIDIA GTX 970M (Maxwell, GM204M) drives the external HDMI-A-1 monitor; nvidia-580xx 580.178.04, modeset=Y - Both outputs are 1920x1080@60 Symptoms are the same as described: windows and the panel flicker on the external monitor only. The wallpaper itself does not flicker. If a window sits across both screens, only the part on the external screen flickers. Regression isolation: - Started with the 6.7.5 -> 6.8.1 upgrade. The NVIDIA driver was unchanged (580.178.04 had been installed for two months). - That update also changed the kernel (7.2.9 -> 7.2.11), but it still flickers on the unchanged LTS kernel 6.18.55, so the kernel is not the cause. Workaround: KWIN_DRM_NO_AMS=1 (via ~/.config/environment.d/) fixes the flicker completely on kernel 7.2.11. supportInformation confirms "Atomic Mode Setting on GPU 0/1: false". I haven't tested KWIN_DRM_FORCE_GL_FINISH_MGPU_COPY=1 or KWIN_ALLOW_VULKAN_MGPU_COPIES=1 yet. Possibly related: on the first boot after the update (kernel 7.2.11, AMS on), KFENCE caught 3 use-after-free reads in the NVIDIA driver, all in kwin_wayland's PRIME import path. It did not show up on the LTS kernel (where it still flickered) or with KWIN_DRM_NO_AMS=1. So it's probably a separate NVIDIA driver bug that is triggered by the new mgpu path rather than the cause of the flicker: BUG: KFENCE: use-after-free read in nv_dma_release_sgt+0x3a/0x70 [nvidia] nv_dma_release_sgt+0x3a/0x70 [nvidia] [... nvidia rm internals ...] rm_ioctl+0x7f/0x3d0 [nvidia] nvidia_unlocked_ioctl+0x5e6/0x7b0 [nvidia] __x64_sys_ioctl+0x144/0x310 allocated by task 767 (kwin_wayland): nv_drm_gem_prime_import_sg_table+0x2e/0xb0 [nvidia_drm] drm_gem_prime_import_dev+0xc8/0x100 drm_gem_prime_fd_to_handle+0x7d/0x1c0 drm_ioctl+0x319/0x4e0 freed by task 767 (kwin_wayland), 1.3 ms later: nv_dma_release_sgt+0x2d/0x70 [nvidia] [... same rm call chain ...] Comm: kwin_wayland Tainted: P IOE 7.2.11-1-cachyos Happy to provide the full trace and KWin supportInformation if needed. -- You are receiving this mail because: You are watching all bug changes.
