https://bugs.kde.org/show_bug.cgi?id=525746

Rikard Nyquist <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REPORTED                    |CONFIRMED
     Ever confirmed|0                           |1

--- Comment #10 from Rikard Nyquist <[email protected]> ---
I had an issue with the cpu usage and my AI came to the conclusion that this
bug might be what is causing it. This is the last response I got, it might help
or not:

"
I can reproduce this on Fedora 44 with plasma-workspace 6.7.5 / kwin 6.7.5
(Wayland session, AMD GPU), with the same "There are no outputs - creating
placeholder screen" flood in the journal when the TV (connected via HDMI) is
switched off and all other outputs are gone.
I captured profiling data during an attack:
plasmashell (~90-100% CPU): perf shows the Mesa GL thread (plasmashel:gl0)
stuck in a loop:
glthread_unmarshal_batch → st_Clear → st_update_framebuffer_state →
dri_st_framebuffer_validate → update_buffers_if_needed →
wl_display_roundtrip_queue → wl_display_dispatch_queue
(~40% of samples in this path)
kwin_wayland (~60% CPU): main thread busy in the event loop — perf shows
qt_safe_poll/ppoll/sock_poll/unix_poll/__pollwait dominating, i.e. it's
drowning in socket traffic rather than computing.
strace counts during an attack (~10 s):
kwin_wayland: 4,772,319 recvmsg / 2,387,075 sendmsg (with 2,386,155 EAGAIN
recvmsg)
plasmashell: 505,038 recvmsg / 252,519 sendmsg (252,519 EAGAIN)
i.e. the plasma GL thread is doing tens of thousands of wl_display_roundtrips
per second against kwin while no real outputs exist.
Journal at attack start: every KDE client logs "There are no outputs - creating
placeholder screen" at the same second, plasmashell logs "requesting unexisting
screen available rect -1", and the HDMI audio sink disappears.
systemctl --user restart plasma-plasmashell.service stops it temporarily, as
others have reported. Happy to provide the full perf reports or perf.data if
useful.
"

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to