Created attachment 152551 [details] dmesg with drm debugging SUMMARY After commit f07d6bd40034372b7176f57de080003c8a612d57, when I move my mouse from my laptop display to my externally-connected HDMI display, the second display goes black. If I have the Display Configuration open, it says that a devices has been removed and it needs to refresh. This video shows what I am experiencing: https://youtu.be/ncIfA-Q5Mxw Oddly, I am not able to reproduce this with just kwin_wayland running in a TTY. STEPS TO REPRODUCE 1. Log in to plasma Wayland with any build containing f07d6bd40034372b7176f57de080003c8a612d57 2. Move mouse from primary display onto USB-C connected HDMI display 3. OBSERVED RESULT Display "disconnects". EXPECTED RESULT Mouse moves smoothly from one display to the next. SOFTWARE/OS VERSIONS Linux/KDE Plasma: 5.19.8-200.fc36.x86_64 (available in About System) KDE Plasma Version: 5.25.5 KDE Frameworks Version: 5.98.0 Qt Version: 5.15.6 ADDITIONAL INFORMATION This is a dual GPU setup, with the laptop's internal display connected to an AMD 680M and the HDMI display connected over USB-C (evdi driver). Attached file contains the DRM logs, collected as described here: https://invent.kde.org/plasma/kwin/-/wikis/Debugging-DRM-issues
Can you also get kwin's debug logs please, i.e. when kwin runs with QT_LOGGING_RULES="kwin_*.debug=true" envvar?
Created attachment 152568 [details] startplasma_wayland with QT_LOGGING_RULES="kwin_*.debug=true" I don't get anything interesting-looking in the log file. I switched to a TTY, ran `export QT_LOGGING_RULES="kwin_*.debug=true"` then `startplasma-wayland > ~/startplasma-wayland.log 2>&1`, reproduced my issue, and logged out. File attached.
Relevant lines from the drm log: [ 4784.121765] evdi: [I] (card0) Closed by Task 131081 (BusImpl_0000559) of process 131079 (DisplayLinkMana) [ 4784.121766] [drm:drm_mode_object_put.part.0] OBJ ID: 43 (3) [ 4784.121768] evdi: [I] (card0) Disconnected from Task 131081 (BusImpl_0000559) of process 131079 (DisplayLinkMana) [ 4784.121710] [drm:amdgpu_get_vblank_counter_kms [amdgpu]] crtc 0: dist from vblank start -1641 [ 4784.121771] evdi: [I] (card0) Removing i2c adapter bus number 1 [ 4784.121805] evdi evdi.0: [drm:check_connector_changed] [CONNECTOR:37:DVI-I-1] status updated from connected to disconnected So without more information I am relatively certain this is a driver bug. I think the correct place to report that is https://github.com/DisplayLink/evdi/issues Until that's fixed, you can try forcing a software cursor as a workaround: https://invent.kde.org/plasma/kwin/-/wikis/Environment-Variables#kwin_force_sw_cursor
(In reply to Zamundaaa from comment #3) > So without more information I am relatively certain this is a driver bug. I > think the correct place to report that is > https://github.com/DisplayLink/evdi/issues > > Until that's fixed, you can try forcing a software cursor as a workaround: > https://invent.kde.org/plasma/kwin/-/wikis/Environment- > Variables#kwin_force_sw_cursor Sorry for the long delay in responding. The KWIN_FORCE_SW_CURSOR is working for me. I no longer have to live in fear of system updates! Do you have a suggestion for what I should tell the EVDI people?
Just tell them it started happening when KWin started using gbm for the cursor. You can also link the issue you create here, then I'll enable notifications there and be available for questions if they have any.
*** Bug 459085 has been marked as a duplicate of this bug. ***
Issue opened here: https://github.com/DisplayLink/evdi/issues/388 I have replaced my displaylink dock because it was causing too many headaches but I still have it on my desk so I can test. Someone who is more impacted (Alexander Reinholdt, if you're following this bug, for instance) may want to take point on responding to the EVDI bug if the developers on that side respond.
(In reply to Simon Redman from comment #7) > Issue opened here: https://github.com/DisplayLink/evdi/issues/388 > > I have replaced my displaylink dock because it was causing too many > headaches but I still have it on my desk so I can test. Someone who is more > impacted (Alexander Reinholdt, if you're following this bug, for instance) > may want to take point on responding to the EVDI bug if the developers on > that side respond. Sure, I will track that bug report and will give answers if requested. Thanks for opening the bug report with the EVDI developers.