CVE-2022-50844

What it actually does

Type mismatch between expected enum PP_OD_DPM_TABLE_COMMAND and u32 for the second parameter of the odn_edit_dpm_table callback in struct amd_pm_funcs, causing Clang CFI to reject the function pointer assignment at compile time and potentially trigger kernel panic or thread termination at runtime on CONFIG_CFI_CLANG kernels.

Why it matters to the business

Systems using AMD GPUs on Linux kernels built with Clang CFI may experience unexpected crashes or instability during power management operations, increasing downtime risk for graphics workloads; low exploitability reduces immediate threat but affects build and stability pipelines.

How an attack unfolds

An attacker with local access would trigger the indirect call through the mismatched function pointer in the AMD power management path (smu_od_edit_dpm_table or pp_odn_edit_dpm_table); CFI validation failure then causes kernel panic or killed thread rather than controlled code execution.

How to fix it

Update to Linux kernel commit f9084e9930db562bdcd47fa199a66fb45e16dab5 or later that corrects the prototype in struct amd_pm_funcs and pp_odn_edit_dpm_table to use enum PP_OD_DPM_TABLE_COMMAND; rebuild kernel with consistent types across amd_pm_funcs, pptable_funcs, and pp_hwmgr_func implementations.

How to confirm the fix worked

Recompile kernel with CONFIG_CFI_CLANG and verify absence of -Wincompatible-function-pointer-types-strict warnings for amdgpu_smu.c:3008; confirm runtime stability during DPM table edits on affected hardware.

Typical attack path

1. Obtain local code execution or user context on affected AMD GPU system
2. Invoke DPM table edit operations that resolve to the mismatched callback
3. Trigger CFI check failure during indirect call validation
4. Result in kernel panic or process termination

How to detect it

Review kernel logs for CFI-related panics or warnings referencing amdgpu_smu.c or odn_edit_dpm_table; examine dmesg and journalctl for function pointer type errors during module load or GPU power events.

What your security team should do

• Prioritize kernel upgrade for any Clang-built CFI kernels running AMD GPU drivers
• Test power management paths post-update before broad deployment
• Monitor for related upstream backports in stable trees