critical bug amdgpu hard reboot after amd-smi reset --gpureset on Strix Point (Radeon 890M)

critical bug amdgpu hard reboot after amd-smi reset --gpureset on Strix Point (Radeon 890M)

[Harris Landgarten]

Hello,

System:

Host: beelink2

Kernel: 7.0.0-rc2-x86_64 #1 SMP PREEMPT_DYNAMIC Sun Mar 1 19:54:38 EST 2026

CPU/GPU: AMD Ryzen AI 9 HX 370 / Radeon 890M (Strix Point)

GPU PCI ID: 1002:150e

Firmware: Beelink T305 BIOS

sys-kernel/linux-firmware-20260221

dev-util/amdsmi-7.1.1

Kernel configuration: PREEMPT_DYNAMIC

Problem:

Running the following command causes the system to hard crash and immediately reboot:

amd-smi reset --gpureset
This behaves like a sudden power loss. In some cases the machine will not restart cleanly until power is physically removed.

This issue has been occurring for some time on 6.18 and 6.19 kernels, but with 7.0-rc2 I was able to capture the failure sequence in the journal before the crash.

Reproduction:

Boot normally into GNOME Wayland.

Run the command as root:

amd-smi reset --gpureset
The GPU reset begins but the system crashes shortly afterward.

Journal excerpt:

Mar 01 20:04:24 beelink2 doas[7188]: harrisl ran command amd-smi reset --gpureset as root from /home/harrisl
Mar 01 20:04:24 beelink2 kernel: amdgpu 0000:c5:00.0: GPU reset begin!. Source:  5
Mar 01 20:04:24 beelink2 kernel: amdgpu 0000:c5:00.0: MODE2 reset
Mar 01 20:04:24 beelink2 kernel: amdgpu 0000:c5:00.0: GPU reset succeeded, trying to resume
Mar 01 20:04:24 beelink2 kernel: amdgpu 0000:c5:00.0: [drm] PCIE GART of 512M enabled (table at 0x0000008000900000).
Mar 01 20:04:24 beelink2 kernel: amdgpu 0000:c5:00.0: SMU is resuming...
Mar 01 20:04:24 beelink2 kernel: amdgpu 0000:c5:00.0: SMU is resumed successfully!
Mar 01 20:04:28 beelink2 kernel: amdgpu 0000:c5:00.0: SMU: No response msg_reg: 31 resp_reg: 0
Mar 01 20:04:28 beelink2 kernel: in params:00000000
Mar 01 20:04:28 beelink2 kernel: amdgpu 0000:c5:00.0: Failed to power ungate VPE!
Mar 01 20:04:28 beelink2 kernel: amdgpu 0000:c5:00.0: [drm] *ERROR* DPM enable vpe failed, ret = -62.
Mar 01 20:04:30 beelink2 kernel: amdgpu 0000:c5:00.0: failed to load ucode VPE(0x42)
Mar 01 20:04:30 beelink2 kernel: amdgpu 0000:c5:00.0: psp gfx command LOAD_IP_FW(0x6) failed and response status is (0x0)
Mar 01 20:04:30 beelink2 kernel: amdgpu 0000:c5:00.0: resume of IP block <vpe_v6_1> failed -22
Mar 01 20:04:30 beelink2 kernel: amdgpu 0000:c5:00.0: GPU reset end with ret = -22
Mar 01 20:04:32 beelink2 kernel: amdgpu 0000:c5:00.0: SMU: No response msg_reg: 31 resp_reg: 0
Mar 01 20:04:32 beelink2 kernel: in params:00000000
Mar 01 20:04:32 beelink2 kernel: amdgpu 0000:c5:00.0: Failed to power gate VCN instance 0!
Mar 01 20:04:32 beelink2 kernel: amdgpu 0000:c5:00.0: [drm] *ERROR* DPM disable vcn failed, ret = -62.
Immediately after this sequence the system performs a hard reset.

Firmware files present:


/lib/firmware/amdgpu/vpe_6_1_0.bin
/lib/firmware/amdgpu/vpe_6_1_1.bin
/lib/firmware/amdgpu/vpe_6_1_3.bin
Questions:

Is this a known issue with amd-smi --gpureset on Strix Point / Radeon 890M?

Does the failure of the VPE resume (vpe_v6_1) during the MODE2 reset sequence indicate a firmware issue or driver bug?

If needed I can provide:

full dmesg

lspci -nnk

complete journal logs from boot to crash

kernel config

Thanks,

Harris Landgarten