**Correction to comment #2**

The 2026-07-20 note that HWE **7.0.0-28** + raw DRI3 had “zero” further
hangs after a 3-day soak is **no longer accurate**. After that soak
window, the same class of hang returned on **7.0.0-28-generic** with
Mesa still **25.2.8-0ubuntu0.24.04.2** (no `LIBGL_DRI3_DISABLE`).
Several kwin_x11 D-Bus wedges occurred from late July through 2026-08-03
(idle-blocked compositor; Xorg stays healthy; recovery = SIGKILL +
restart kwin).

Comment #2’s notification-storm PASS on 7.0 still stands; day-to-day
desktop use (especially **screen lock / DPMS**) continues to hit the
race.

**Environment (unchanged package-wise)**
- Ubuntu 24.04.4, X11, Plasma 5.27 / kwin-x11 5.27.11
- mesa / libglx-mesa0 **25.2.8-0ubuntu0.24.04.2**
- kernel **7.0.0-28-generic** (HWE)
- Dell Precision 5690, Meteor Lake iGPU (Intel Arc MTL), compositing via GLX

**New evidence — kwin_x11 blocking variant (2026-08-03)**

Captured live with gdb while `org.kde.KWin` D-Bus was hung. CPU counters
flat over 3s (blocked, not spinning). Main thread:

```
#0  poll (timeout=-1)
#1  … libxcb …
#2  xcb_wait_for_special_event
#3–#5  … libgallium-25.2.8 …
#6  loader_dri3_query_buffer_age   ← Mesa 25.2.8 DRI3 loader
#7  … libGLX_mesa.so.0 …
#9  KWin::Compositor::composite
#10 KWin::X11Compositor::composite
#12 KWin::RenderLoop::frameRequested
#13 KWin::RenderLoopPrivate::dispatch
… QTimer → QEventLoop …
```

Full `thread apply all bt` attached (`gdb-kwin_x11-dri3-query-buffer-
age-20260803.txt`).

This is the same DRI3 Present / special-event wait family as the
original plasmashell `glXSwapBuffers` hang (comment #1), observed here
in `loader_dri3_query_buffer_age` on the kwin composite path.

**Reproducer (reliable on this machine)**

Trigger is **lock the session, then force DPMS off, then wake the
displays**. On this box that is bound to **Meta+Alt+L** (Windows+Alt+L)
via a small helper (PowerDevil’s built-in “Turn Off Screen” is unbound
because on X11 key-release immediately wakes DPMS — KDE #423035).

Equivalent without our shortcut (run from an already-unlocked session):

```bash
# 1) lock (greeter starts)
qdbus org.freedesktop.ScreenSaver /ScreenSaver Lock

# 2) blank outputs shortly after (race window)
sleep 1
xset dpms force off

# 3) wake with mouse or key
# Expected on this machine within seconds to ~1 minute:
#   - lock greeter is alive (Howdy/fingerprint/password path may run)
#   - display shows a frozen / ghost frame (often last unlocked desktop)
#   - org.kde.KWin supportInformation / D-Bus times out
#   - kwin_x11 stuck as in the attached backtrace
```

Related variant that also led to the same ghost-lock phenotype: **DPMS-
off only** (no lock), then wait for Plasma **Autolock** (here Timeout=10
minutes) so lock runs while outputs are already off.

**Asks**
1. Please re-evaluate backporting the Mesa **26.1.5** `dri3: Increment 
draw->send_sbc after waiting for last presentation` fix (or current equivalent) 
to noble’s **25.2.x**.
2. Happy to capture `LIBGL_DEBUG` / present traces or retest a proposed package 
if useful.


** Attachment added: "kwin_x11 gdb thread apply all bt — 
loader_dri3_query_buffer_age hang 2026-08-03"
   
https://bugs.launchpad.net/bugs/2161079/+attachment/5988814/+files/gdb-kwin_x11-dri3-query-buffer-age-20260803.txt

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2161079

Title:
  kwin_x11 and plasmashell hang forever in DRI3 glXSwapBuffers after HWE
  kernel 6.17 update (Meteor Lake, X11) — fix candidate in Mesa 26.1.5

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/mesa/+bug/2161079/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to