CVE-2021-47550

What it actually does

Missing kobject_put call on kobject_init_and_add failure in amdgpu_get_xgmi_hive leads to reference count leak and gradual kernel memory consumption.

Why it matters to the business

On AMD GPU systems using XGMI, repeated triggering of the error path can exhaust kernel memory, degrading performance or causing denial of service; low urgency given stable low EPSS and absence from KEV.

How an attack unfolds

A local attacker with low privileges repeatedly invokes amdgpu_get_xgmi_hive under conditions where kobject_init_and_add fails, accumulating leaked kobjects until memory pressure occurs.

How to fix it

Update to a Linux kernel containing the fix for CVE-2021-47550; apply the patch that adds the missing kobject_put in the failure path of amdgpu_get_xgmi_hive.

How to confirm the fix worked

Confirm kernel source contains kobject_put after failed kobject_init_and_add in amdgpu_get_xgmi_hive; run memory leak detection tools or monitor /proc/meminfo under load on AMD XGMI hardware.

Typical attack path

1. Obtain local user access on affected Linux kernel with AMDGPU XGMI support
2. Trigger error path in amdgpu_get_xgmi_hive by inducing kobject_init_and_add failure
3. Repeat to accumulate unreleased kobject allocations
4. Observe system-wide memory exhaustion or OOM conditions

How to detect it

Review kernel logs for amdgpu XGMI hive initialization errors; inspect memory usage trends via slabtop or kmemleak on systems running vulnerable kernels.

What your security team should do

• Prioritize kernel upgrade on systems with AMD GPUs and XGMI enabled
• Monitor for unusual kernel memory growth on affected hardware
• Apply defense-in-depth by restricting local unprivileged access where possible