[BUG] amdgpu resume of IP block <smu> failed -62 after 7.1.1-2 → 7.1.2-3 (RX 9070 XT), suspend never wakes

[Madscientist351]

Pre-flight checklist

• I have searched existing issues and this is not a duplicate.
• I have read the Contributing Guidelines.
• I have verified the issue is reproducible with the latest available CachyOS kernel.
• I have tried to reproduce the issue on Arch Linux's linux kernel.

Upstream / vanilla kernel check

I have not tested with a vanilla/upstream kernel

Kernel variant

linux-cachyos (EEVDF, Clang)

System information (cachyos-bugreport.sh)

https://paste.cachyos.org/p/f2dafa6.log

Manual system information (if cachyos-bugreport.sh is unavailable)

Bug description

Reposting here per direct request of Peter Jung (ptr1337) via Reddit.

System

Kernel: 7.1.1-2-cachyos x86_64 (was 7.1.2-3 at time of failure; downgraded for testing, result below)
Desktop: KDE Plasma 6.7.2 / wayland / kwin_wayland
Distro: CachyOS, Arch Linux base

Machine: Gigabyte X870E AORUS PRO, Firmware: UEFI American Megatrends v: F11 (02/05/2026)

CPU: AMD Ryzen 7 7800X3D (8-core, Zen 4)

Graphics:
  Device-1: AMD Navi 48 [Radeon RX 9070 XT/9070 GRE] driver: amdgpu v: kernel
    RDNA-4, pcie speed: 32 GT/s, lanes: 16, chip-ID: 1002:7550, bus-ID: 03:00.0
  Display: wayland, compositor: kwin_wayland, driver: gpu: amdgpu
  Monitors: DP-1 (primary, 2560x1440@165Hz) + HDMI-A-1 (1920x1080@60Hz)
  OpenGL: v4.6, mesa: 26.1.4-arch3.1, renderer: AMD Radeon RX 9070 XT (radeonsi gfx1201 ACO)
  Vulkan: v1.4.350

Sensors: gpu temp: 57.0 C, gpu fan: 0 (zero-RPM idle curve, expected/normal)

Power: states: freeze,mem,disk / suspend: deep (default) / hibernate: platform

Summary

Suspend (idle-triggered, PowerDevil "Sleep after 120 min") worked normally prior to the 2026-07-01 update. Fans/AIO pump/RGB shut off as expected, tap-to-wake worked. After updating linux-cachyos 7.1.1-2 → 7.1.2-3 (same transaction also bumped mesa 3.26.1.2 → 3.26.1.3), suspend still triggers but the system never fully wakes — CPU/fans keep running, but the GPU/displays are dead until a hard reboot.

Log evidence (journalctl -b -1 -k)

PM: suspend entry (deep)
amdgpu: Failed to enable requested dpm features!
amdgpu: Failed to setup smc hw!
amdgpu: resume of IP block <smu> failed -62
amdgpu_device_ip_resume failed (-62)
PM: suspend exit
PM: suspend entry (s2idle)
[cascading failures follow: suspend of IP block <smu> failed -22, GPU reset attempted,
 sdma ring test failed (-110), GPU Recovery Failed: -62]

Full suspend is entered (deep/S3), but resume fails at the SMU firmware handshake before the display controller is ever reached in the resume sequence — which is consistent with why the monitors specifically never recover.

What's been ruled out

• Mesa 3.26.1.3 — bumped same transaction, but mesa has no code path in kernel-space suspend/resume; ruled out on architectural grounds
• OpenLinkHub udev rules (/etc/udev/rules.d/99-openlinkhub.rules, installed the day before) — inspected directly, contains only USB device-permission grants (MODE=0660, GROUP=openlinkhub) for Corsair vendor ID 1b1c, no power-management directives. Ruled out by inspection, not just timing.
• linux-firmware-amdgpu — last updated 5 days prior to the issue first appearing, so unlikely but not fully excluded

Status / what I still need to confirm

• Whether 7.1.1-2 (currently testing) resolves it cleanly over a real 2hr idle cycle — not yet confirmed pass/fail
• Whether linux-cachyos-lts also reproduces this
• Reproducibility via manual systemctl suspend vs. only idle-triggered (haven't isolated this yet)

Note: the current test environment is kernel 7.1.1-2 + mesa 26.1.4 (mesa updated again since the original failure, which was on mesa 26.1.3). Not a clean single-variable rollback of the original working state (kernel 7.1.1-2 + mesa 26.1.2) — flagging so a pass result isn't over-read as fully isolating the kernel as sole cause.

Will update this issue once the 7.1.1-2 test completes. Posting now per ptr1337's request on r/cachyos.

Steps to reproduce

My understanding is this is a long standing bug with linux

Expected behavior

Computer to enter deep sleep state properly

Actual behavior

Computer doesn't enter deep sleep state properly

Logs / stack traces

Additional system information

Thanks for all the hard work you do! CachyOS is my first daily drive of Linux and so far it's been really fantastic!

System:
Kernel: 7.1.1-2-cachyos arch: x86_64 bits: 64 compiler: clang v: 22.1.6
clocksource: tsc
Desktop: KDE Plasma v: 6.7.2 tk: Qt v: N/A wm: kwin_wayland vt: 2 dm: N/A
Distro: CachyOS base: Arch Linux
Machine:
Type: Desktop System: Gigabyte product: X870E AORUS PRO v: N/A
serial:
Mobo: Gigabyte model: X870E AORUS PRO serial:
uuid: Firmware: UEFI vendor: American Megatrends LLC.
v: F11 date: 02/05/2026
Battery:
Device-1: corsair-void-4-battery model: Corsair CORSAIR VOID ELITE Wireless
Gaming Dongle serial: N/A charge: normal status: full
CPU:
Info: 8-core model: AMD Ryzen 7 7800X3D bits: 64 type: MT MCP smt: enabled
arch: Zen 4 rev: 2 cache: L1: 512 KiB L2: 8 MiB L3: 96 MiB
Speed (MHz): avg: 3905 min/max: 426/5053 boost: enabled cores: 1: 3905
2: 3905 3: 3905 4: 3905 5: 3905 6: 3905 7: 3905 8: 3905 9: 3905 10: 3905
11: 3905 12: 3905 13: 3905 14: 3905 15: 3905 16: 3905 bogomips: 134141
Flags-basic: avx avx2 ht lm nx pae sse sse2 sse3 sse4_1 sse4_2 sse4a ssse3
Graphics:
Device-1: Advanced Micro Devices [AMD/ATI] Navi 48 [Radeon RX 9070/9070
XT/9070 GRE] vendor: Tul / PowerColor driver: amdgpu v: kernel
arch: RDNA-4 pcie: speed: 32 GT/s lanes: 16 ports: active: DP-1,HDMI-A-1
empty: DP-2,DP-3,Writeback-1 bus-ID: 03:00.0 chip-ID: 1002:7550
class-ID: 0300
Display: wayland server: Xwayland v: 24.1.12 compositor: kwin_wayland
driver: gpu: amdgpu d-rect: 4480x2520 display-ID: 0
Monitor-1: DP-1 pos: primary,top-left model: Dell G2724D serial:
res: mode: 2560x1440 hz: 165 scale: 100% (1) dpi: 109
size: 596x335mm (23.46x13.19") diag: 684mm (26.9") modes: max: 2560x1440
min: 720x400
Monitor-2: HDMI-A-1 pos: bottom-r model: Acer XFA240 serial: res:
mode: 1920x1080 hz: 60 scale: 100% (1) dpi: 92
size: 531x299mm (20.91x11.77") diag: 609mm (24") modes: max: 1920x1080
min: 720x400
API: EGL v: 1.5 hw: drv: amd radeonsi platforms: device: 0 drv: radeonsi
device: 1 drv: swrast gbm: drv: radeonsi surfaceless: drv: radeonsi wayland:
drv: radeonsi x11: drv: radeonsi
API: OpenGL v: 4.6 vendor: amd mesa v: 26.1.4-arch3.1 glx-v: 1.4
direct-render: yes renderer: AMD Radeon RX 9070 XT (radeonsi gfx1201 ACO
DRM 3.64 7.1.1-2-cachyos) device-ID: 1002:7550 display-ID: :0.0
API: Vulkan v: 1.4.350 layers: 9 surfaces: N/A device: 0
type: discrete-gpu hw: amd driver: mesa radv device-ID: 1002:7550
Info: Tools: api: clinfo, eglinfo, glxinfo, vulkaninfo
de: kscreen-console,kscreen-doctor gpu: lact wl: wayland-info
x11: xdpyinfo, xprop, xrandr
Audio:
Device-1: Advanced Micro Devices [AMD/ATI] Navi 48 HDMI/DP Audio
driver: snd_hda_intel v: kernel pcie: speed: 32 GT/s lanes: 16
bus-ID: 03:00.1 chip-ID: 1002:ab40 class-ID: 0403
Device-2: Advanced Micro Devices [AMD] Ryzen HD Audio vendor: Gigabyte
driver: snd_hda_intel v: kernel pcie: speed: 16 GT/s lanes: 16
bus-ID: 7b:00.6 chip-ID: 1022:15e3 class-ID: 0403
Device-3: Corsair VOID ELITE Wireless Gaming Dongle
driver: hid-corsair-void,snd-usb-audio,usbhid type: USB rev: 1.1
speed: 12 Mb/s lanes: 1 bus-ID: 1-2:2 chip-ID: 1b1c:0a51 class-ID: 0300
API: ALSA v: k7.1.1-2-cachyos status: kernel-api
Server-1: sndiod v: N/A status: off
Server-2: JACK v: 1.9.22 status: off
Server-3: PipeWire v: 1.6.7 status: active with: 1: pipewire-pulse
status: active 2: wireplumber status: active 3: pipewire-alsa type: plugin
Network:
Device-1: Qualcomm WCN785x Wi-Fi 7 320MHz 2x2 [FastConnect 7800]
vendor: Foxconn driver: ath12k_wifi7_pci v: N/A pcie: speed: 8 GT/s lanes: 1
bus-ID: 0f:00.0 chip-ID: 17cb:1107 class-ID: 0280
IF: wlan0 state: down mac:
Device-2: Realtek RTL8125 2.5GbE vendor: Gigabyte driver: r8169 v: kernel
pcie: speed: 5 GT/s lanes: 1 port: e000 bus-ID: 10:00.0 chip-ID: 10ec:8125
class-ID: 0200
IF: enp16s0 state: up speed: 1000 Mbps duplex: full mac:
Device-3: Intel Ethernet 10-Gigabit X540-AT2 driver: ixgbe v: kernel pcie:
speed: 5 GT/s lanes: 4 port: N/A bus-ID: 12:00.0 chip-ID: 8086:1528
class-ID: 0200
IF: enp18s0f0 state: down mac:
Device-4: Intel Ethernet 10-Gigabit X540-AT2 driver: ixgbe v: kernel pcie:
speed: 5 GT/s lanes: 4 port: N/A bus-ID: 12:00.1 chip-ID: 8086:1528
class-ID: 0200
IF: enp18s0f1 state: down mac:
Bluetooth:
Device-1: Foxconn / Hon Hai driver: btusb v: 0.8 type: USB rev: 1.1
speed: 12 Mb/s lanes: 1 bus-ID: 1-6:4 chip-ID: 0489:e10d class-ID: e001
Report: btmgmt ID: hci0 rfk-id: 0 state: down bt-service: disabled
rfk-block: hardware: no software: yes address: N/A
Drives:
Local Storage: total: 2.28 TiB used: 1.24 TiB (54.5%)
ID-1: /dev/nvme0n1 vendor: Samsung model: SSD 990 EVO 2TB size: 1.82 TiB
speed: 126 Gb/s lanes: 4 tech: SSD serial: fw-rev: 0B2QKXJ7
temp: 50.9 C scheme: GPT
ID-2: /dev/sda vendor: Samsung model: SSD 860 EVO 500GB size: 465.76 GiB
speed: 6.0 Gb/s tech: SSD serial: fw-rev: 4B6Q scheme: GPT
ID-3: /dev/sdb vendor: PNY model: USB 2.0 FD size: 7.55 GiB type: USB
rev: 2.0 spd: 480 Mb/s lanes: 1 tech: N/A serial: fw-rev: 1100
scheme: GPT
Partition:
ID-1: / size: 465.26 GiB used: 63.47 GiB (13.6%) fs: btrfs dev: /dev/sda2
ID-2: /boot/efi size: 511 MiB used: 664 KiB (0.1%) fs: vfat dev: /dev/sda1
ID-3: /home size: 465.26 GiB used: 63.47 GiB (13.6%) fs: btrfs
dev: /dev/sda2
ID-4: /var/log size: 465.26 GiB used: 63.47 GiB (13.6%) fs: btrfs
dev: /dev/sda2
ID-5: /var/tmp size: 465.26 GiB used: 63.47 GiB (13.6%) fs: btrfs
dev: /dev/sda2
Swap:
ID-1: swap-1 type: zram size: 30.95 GiB used: 1.7 MiB (0.0%) priority: 100
dev: /dev/zram0
Sensors:
System Temperatures: cpu: 45.2 C mobo: N/A gpu: amdgpu temp: 57.0 C
mem: 76.0 C
Fan Speeds (rpm): N/A gpu: amdgpu fan: 0
Info:
Memory: total: 32 GiB note: est. available: 30.95 GiB used: 4.61 GiB (14.9%)
Processes: 405 Power: uptime: 52m states: freeze,mem,disk suspend: deep
wakeups: 0 hibernate: platform Init: systemd v: 261 default: graphical
Packages: 1448 pm: pacman pkgs: 1439 pm: flatpak pkgs: 9 Compilers:
clang: 22.1.6 gcc: 16.1.1 Shell: fish v: 4.8.0 running-in: konsole
inxi: 3.3.40

[Madscientist351]

Rolling the kernel version back to 7.1.1-2 appears to have resolved the issue

[ptr1337]

Can you try to revert CachyOS/linux@68c43a6 ?

[Madscientist351]

Can you try to revert CachyOS/linux@68c43a6 ?

Honestly, I haven't built a kernel from source before. Is there any chance you could point me to a test package/branch with that commit already reverted, or a PKGBUILD I could just run through makepkg? Happy to test whatever you send, just want to make sure I don't botch a manual revert-and-build.

[1Naim]

https://paste.cachyos.org/p/528c9aa add this patch to kernel manager or the source=() array in PKGBUILD if u want to use makepkg.

[Madscientist351]

I'll get to work on getting 7.1.2-3 compiled with that commit added. Also an update for similar symptoms on 7.1.1-2 but what appears to be an entirely different cause. Claude AI helped me root through the system logs and prepared this summary based on our findings.

Update: recurrence on 7.1.1-2 — appears to be a separate bug with the same symptom

Wanted to flag this rather than assume it's the same issue, since the log evidence points somewhere different.

Context: After confirming 7.1.1-2 fixed the original SMU resume issue for ~2 clean days, the "monitors dead, system still running" symptom recurred last night — but on 7.1.1-2, not 7.1.2-3.

Checked first: is this the same SMU resume bug?
No. journalctl -b -1 -k for this occurrence shows:

• No PM: suspend entry line at all — full system suspend was never triggered
• No resume of IP block <smu> failed anywhere in the log
  So this is not a recurrence of the original issue — different trigger, different mechanism, same visible symptom (dead display, box still running, requires hard reboot).

What actually shows up instead:

1. amdgpu runtime PM (autosuspend) cycling every ~2 min, unrelated to system suspend:

dm_suspend.llvm.[...]+0x53/0x480 [amdgpu]
amdgpu_ip_block_suspend+0x27/0x70 [amdgpu]
amdgpu_device_ip_suspend_phase1+0x200/0x410 [amdgpu]

followed later by a pattern shift to bare __pm_runtime_suspend+0x63/0xc0 calls every ~30 sec.

2. A recurring kernel WARNING, timestamps exactly matching the __pm_runtime_suspend calls above:

------------[ cut here ]------------
WARNING: drivers/gpu/drm/ttm/ttm_bo_util.c:745 at ttm_bo_move_memcpy+0x258/0x2c0 [ttm], CPU#2: kwin_wayla:cs/0/1449
Tainted: [S]=CPU_OUT_OF_SPEC, [W]=WARN
CPU: 2 UID: 1000 PID: 1449 Comm: kwin_wayla:cs/0 Tainted: G S W 7.1.1-2-cachyos #1 PREEMPT(full)
Hardware name: Gigabyte Technology Co., Ltd. X870E AORUS PRO/X870E AORUS PRO, BIOS F11 02/05/2026
Call Trace:
 amdgpu_bo_move+0xcdf/0x1060 [amdgpu]
 ttm_bo_handle_move_mem+0x13f/0x180 [ttm]
 ttm_bo_validate+0x115/0x270 [ttm]
 amdgpu_cs_bo_validate+0xe4/0x1d0 [amdgpu]
 amdgpu_cs_ioctl+0x1646/0x2570 [amdgpu]

Triggered by kwin_wayla:cs/0 (KWin's compositor submission thread). Checked against current upstream ttm_bo_move_memcpy — the only WARN_ON in that function is WARN_ON(!src_mem), so this fires specifically when a buffer object is handed to the memcpy-move path with no valid source memory resource attached (returns -EINVAL, does not itself crash).

Working theory (not confirmed): a race between amdgpu's runtime PM trying to suspend/evict the GPU and KWin's compositor concurrently submitting buffer moves — one side tearing down/reassigning a BO's memory resource while the other is still trying to operate on it. Recurrence lines up with having GPU-accelerated apps open (Discord, confirmed not autostart but present after the hard reset, so was genuinely open at crash time) alongside normal desktop compositing.

Honest caveat: the WARN_ON itself returns an error and lets the kernel continue — it's not a crash on its own. Haven't yet confirmed whether this warning is directly causal to the display/GPU death, or a correlated-but-separate symptom of the same underlying PM/compositor race. Still isolating what happens in the log at the actual moment the monitors go dark.

Also noted, probably unrelated but flagging for completeness: Tainted: [S]=CPU_OUT_OF_SPEC — CPU running outside documented spec at some point this boot (common with X870E BIOS defaults / PBO). No known connection to this bug, just visible in the taint flags.

Confirmed system state: kernel 7.1.1-2-cachyos, RX 9070 XT, X870E AORUS PRO, KDE Plasma 6.7.2/Wayland — same system as the original report.

Happy to grab more targeted log windows if useful — let me know what would help narrow this down further.

[Madscientist351]

Can you try to revert CachyOS/linux@68c43a6 ?

Been on the patched linux-cachyos-7.1.2-3 kernel for a couple days now. No issues with the PC failing to sleep/wake normally so far

Noticed an update to kernel version 7.1.3-2. Would you recommend I go ahead an update to the new kernel or remain on the patched 7.1.2-3 for now?