GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel hisilicon/sec2: use-after-free on the request path under heavy load

CVSS 9.8CVE-2026-53055Kernel, userspace & hypervisorcurated

Impact

The sec2 driver kept using the request object after the hardware had already completed the packet and freed it, so a loaded machine can touch freed memory on the submit path. The practical outcome is kernel memory corruption from a local workload that drives the crypto accelerator hard - crash of a node at minimum, with the usual UAF potential for privilege escalation. This only matters on hosts built on HiSilicon Kunpeng server SoCs that expose the SEC2 engine; on an x86 or NVIDIA-based GPU node the driver is not present and the CVE is irrelevant. Note that the NVD vector says AV:N; the described defect is reached by local work submitted to the accelerator, not by a remote packet, so treat the 9.8 as unrepresentative.

Who can reach it

Local user on an affected HiSilicon platform able to submit crypto work that reaches the SEC2 engine (for example through the kernel crypto API), under enough load to race hardware completion. No authentication beyond local access.

What to do

Take the stable kernel fix (the linked git.kernel.org commits replace the freed req with the longer-lived qp_ctx) and reboot each affected node; there is no runtime mitigation short of not using the SEC2 accelerator. Fleets without HiSilicon accelerators need no action. The record names no vendor-fixed version beyond the stable commits.

References

Related entries

All Kernel, userspace & hypervisor entries

This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.