GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel io_uring: registered-buffer folio offset miscalculated for unaligned user pointers

CVSS 7.8CVE-2025-40216Kernel, userspace & hypervisorcurated

Impact

io_uring registered buffers computed the offset of the first page into a coalesced folio with bit-mask logic that assumed the user virtual address was aligned. With an unaligned pointer the wrong offset is used, so I/O can land on the wrong bytes of the mapping - the NVD scoring treats this as local confidentiality, integrity and availability impact. On a GPU node this matters because io_uring is reachable from inside any unprivileged container unless seccomp blocks it, so a tenant process on a shared node is the attacker population. The record is a stable-tree fix with no exploit detail, so treat the ceiling as memory corruption in kernel context rather than a demonstrated escape.

Who can reach it

Local, unprivileged: any process that can call io_uring_setup and register buffers, including a tenant inside a GPU pod when io_uring is not blocked by seccomp. No authentication beyond having code execution on the node.

What to do

Take the stable kernel update that carries the io_uring/rsrc fix (commits linked in the record) and reboot each node; there is no runtime toggle for the bug itself. As a stopgap on multi-tenant GPU nodes, disable io_uring for untrusted workloads - kernel.io_uring_disabled sysctl, or a seccomp profile that denies io_uring_setup - which avoids the reboot until a maintenance window. No fixed release version is stated in the record beyond the listed 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.