Public bug reported:
With two displays connected and active, the machine no longer completes
poweroff or reboot. The displays go dark, but the board never enters S5:
the fans spin up to full speed and the system stays powered on
indefinitely. Only a long press on the power button ends the state.
With a single display connected, poweroff and reboot always complete
normally.
This is a regression. The machine worked correctly for months on the
6.17 kernel series and broke on the day the 7.0 series was installed.
Hardware
GPU: AMD Navi 48 [Radeon RX 9070 XT], PCI ID 1002:7550, board vendor ASUSTeK
PCIe link: 32 GT/s, 16 lanes, bus-ID 03:00.0
Motherboard: ASUS TUF Z390-PLUS GAMING (WI-FI), UEFI 3006 (latest available)
CPU: Intel Coffee Lake (Z390 platform)
Software
Distribution: Linux Mint 22.3 (base: Ubuntu 24.04 noble)
Desktop: Cinnamon 6.6.9, LightDM, X11 (Xorg 21.1.11)
Mesa: 25.2.8-0ubuntu0.24.04.2 (radeonsi, gfx1201, LLVM 20.1.2, DRM 3.64)
Bad kernels: 7.0.0-28, 7.0.0-29, 7.0.0-30, 7.0.0-31 (Ubuntu HWE builds)
Last known good kernel: 6.17.0-35 (Ubuntu HWE build)
Displays
Connector Model Resolution Cable
DP-2 AOC AG273QS8R4 (AGON AG273QZ), 27" 2560x1440 native DP to DP
DP-3 LG (GoldStar) W2353, 23" 1920x1080 DP to HDMI (passive)
HDMI-A-1 Samsung TV, 54.6" 1920x1080 HDMI
Reproducer
Consistent across many attempts:
Any single display, on any port: poweroff completes normally. Always.
AOC AG273QS8R4 plus any second active display: hang. Reproduced with both the
LG monitor and the Samsung TV as the second display.
LG plus Samsung TV, without the AOC: poweroff completes normally (verified
three times).
If the TV has been powered off long enough for the link to be torn down,
poweroff sometimes completes even with the cable attached — consistent with
only one display actually being active at that point.
So the trigger appears to require the AOC panel specifically, in
combination with a second active output. Two active outputs alone are
not sufficient.
Additional observations
Also occurs on reboot, not only on poweroff.
Also occurs with no X server running (systemctl stop lightdm, then systemctl
poweroff from a text console). The display server is not involved.
The systemd shutdown sequence itself completes cleanly. The journal ends with
the expected entries:
systemd[1]: Reached target poweroff.target - System Power Off.
systemd[1]: Shutting down.
systemd-shutdown[1]: Syncing filesystems and block devices.
systemd-shutdown[1]: Sending SIGTERM to remaining processes...
systemd-journald[314]: Received SIGTERM from PID 1 (systemd-shutdow).
systemd-journald[314]: Journal stopped
Nothing is logged after that point, so the hang occurs during or after
the final driver teardown, before the ACPI poweroff takes effect.
Regression window
First occurrence: 2026-07-24, 22:16 local time.
linux-image-7.0.0-28-generic was installed the same day at 09:41 local time
(dpkg log). The machine had been powered off for over a week, was booted that
morning, pulled the update, and hung on the first shutdown that evening.
The previously running series was 6.17. Booting 6.17.0-35 today, with the exact
configuration that hangs on 7.0.0-31, completes poweroff cleanly (verified
three times).
So the regression was introduced somewhere between 6.17 and 7.0.
Ruled out
Kernel parameters (no effect):
amdgpu.runpm=0
amdgpu.aspm=0
Firmware / BIOS:
BIOS updated from 1002 to 3006 (latest) — no change.
ASPM fully disabled in UEFI: PCI Express Native Power Management, PCH DMI ASPM,
ASPM, L1 Substates, DMI Link ASPM Control, PEG-ASPM — all Disabled.
PCI Express Clock Gating disabled as a test — no change.
Hardware / peripheral:
Cable replaced, including a switch from passive DP-to-HDMI to native DP-to-DP
on the AOC — no change.
FreeSync / Adaptive-Sync disabled in the monitor OSD — no change.
Various other monitor OSD settings tried — no change.
PSU is new; power button behaviour is otherwise normal.
Other OS:
Windows on the same machine, same GPU, same monitors and cables, shuts down
cleanly. Monitor firmware, GPU VBIOS and board firmware can therefore be
excluded.
What I can provide
Happy to supply further logs, test specific kernels, or apply patches. A
full bisect between 6.17 and 7.0 is possible if needed, though it would
take some time as each test requires a physical hard power-off.
Appendix: inxi -GSMxx output
System:
Host: maxlinux-System-Product-Name Kernel: 6.17.0-35-generic arch: x86_64
bits: 64 compiler: gcc v: 13.3.0
Desktop: Cinnamon v: 6.6.9 tk: GTK v: 3.24.41 wm: Muffin dm: LightDM
Distro: Linux Mint 22.3 Zena base: Ubuntu 24.04 noble
Machine:
Type: Desktop Mobo: ASUSTeK model: TUF Z390-PLUS GAMING (WI-FI) v: Rev X.0x
part-nu: ASUS_MB_CNL UEFI: American Megatrends v: 3006 date: 10/12/2021
Graphics:
Device-1: AMD Navi 48 [Radeon RX 9070/9070 XT/9070 GRE] vendor: ASUSTeK
driver: amdgpu v: kernel pcie: speed: 32 GT/s lanes: 16 ports:
active: DP-2,DP-3 off: HDMI-A-1 empty: DP-1,Writeback-1 bus-ID: 03:00.0
chip-ID: 1002:7550
Display: x11 server: X.Org v: 21.1.11 with: Xwayland v: 23.2.6 driver: X:
loaded: amdgpu unloaded: fbdev,modesetting,radeon,vesa dri: radeonsi
gpu: amdgpu display-ID: :0 screens: 1
Screen-1: 0 s-res: 4480x1440 s-dpi: 96
Monitor-1: DP-2 mapped: DisplayPort-1 pos: primary,right
model: AOC AG273QS8R4 res: 2560x1440 dpi: 109 diag: 685mm (27")
Monitor-2: DP-3 mapped: DisplayPort-2 pos: left model: LG (GoldStar) W2353
res: 1920x1080 dpi: 96 diag: 587mm (23.1")
Monitor-3: HDMI-A-1 mapped: HDMI-A-0 note: disabled model: Samsung
res: 1920x1080 dpi: 81 diag: 1388mm (54.6")
API: OpenGL v: 4.6 compat-v: 4.5 vendor: amd mesa
v: 25.2.8-0ubuntu0.24.04.2 glx-v: 1.4 direct-render: yes renderer: AMD
Radeon RX 9070 XT (radeonsi gfx1201 LLVM 20.1.2 DRM 3.64
6.17.0-35-generic) device-ID: 1002:7550
API: Vulkan v: 1.3.275 surfaces: xcb,xlib device: 0 type: discrete-gpu
driver: N/A device-ID: 1002:7550
Note: captured while running the last known good kernel (6.17.0-35).
** Affects: linux (Ubuntu)
Importance: Undecided
Status: New
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2167184
Title:
Regression in 7.0 HWE kernel: system hangs at poweroff with two
active displays on Navi 48 (RX 9070 XT)
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2167184/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
