Update with more systematic reproduction and bisection data.

Hardware confirmation (correcting/confirming the original report):
- CPU: AMD Ryzen 5 7535U with Radeon Graphics
- GPU: AMD Rembrandt [Radeon 680M] [1002:1681] — confirms the Rembrandt/
  Radeon 680M identification in the original report is accurate.
- BIOS: updated from 01.08.03 to W78 Ver. 01.09.00 (06/10/2026) since
  original report — issue persists unchanged on the new BIOS.

Kernel bisection (using locally available installed kernels, no external
builds):
- 7.0.0-22-generic: GOOD — confirmed reliably clean across multiple
  separate shutdown/reboot cycles, no kernel parameters or BIOS changes
  needed.
- 7.0.0-29-generic: BAD — hangs with identical symptom.
- 7.0.0-31-generic: BAD — hangs with identical symptom, reproduced
  repeatedly.

=> Regression window narrowed to somewhere between 7.0.0-22 and
   7.0.0-29 (Ubuntu 26.04 HWE kernel point releases).

Workarounds tested on the BAD kernel (7.0.0-31), none resolved the issue:
- BIOS "Pre-boot DMA Protection" (AMD Rembrandt platform feature):
  disabled — no change.
- Kernel parameter amd_iommu=off — no change (confirmed via dmesg/
  /sys/kernel/iommu_groups that AMD-Vi was genuinely not initialized
  during the test, so this was a real test, not a no-op).
- Kernel parameter pcie_aspm=off — no change.

Symptom detail for clarity: after systemd completes the shutdown sequence
cleanly (poweroff.target reached, systemd-journald stops logging as
expected), the machine does not reach true ACPI S5. Symptoms observed
post-"shutdown": power/status LED remains lit, and the keyboard backlight
still visibly reacts to keypresses. This indicates the EC/keyboard
controller domain remains active, not just a cosmetic LED issue. Only a
~10s hard power-button press fully cuts power. Left unattended overnight
unplugged, this previously drained the battery completely; with kernel
7.0.0-22 the battery loss overnight was negligible (~2%), consistent with
a genuine clean poweroff.

Current mitigation: pinned to 7.0.0-22-generic via apt-mark hold pending
a fix. Happy to test further bisection (7.0.0-23 through -28) from the
Ubuntu archive if that would help narrow the regression commit range
further — let me know if that's useful or if the 22/29 boundary is
already enough to work with.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2163015

Title:
  System fails to power off / hangs on shutdown-reboot after upgrade to
  Kubuntu 26.04 (kernel 7.0)

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2163015/+subscriptions


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

Reply via email to