GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel perf: addr_filter_ranges freed without RCU grace period, leaving a use-after-free

CVSS 7.8CVE-2026-93817Kernel, userspace & hypervisorcurated

Impact

perf_event::addr_filter_ranges is read under RCU but was not freed through RCU, so a concurrent reader can touch the freed array - a classic use-after-free in core kernel perf. The record is terse: it states the lifetime bug and the fix, and no exploitation detail. Exposure depends entirely on who may call perf_event_open with address filters; GPU nodes that run fleet-wide profiling or that have relaxed perf_event_paranoid for tenant benchmarking widen that set considerably, and a host-kernel UAF on a shared GPU host is a cross-tenant problem.

Who can reach it

Local user able to call perf_event_open and set address filters. Gated by kernel.perf_event_paranoid and by CAP_PERFMON; not reachable from a pod with perf syscalls blocked.

What to do

Take a stable kernel carrying the RCU-free fix for addr_filter_ranges (commits in the record), which means a kernel update and a reboot of each node. Until then, check kernel.perf_event_paranoid on GPU hosts and do not leave it relaxed for tenant workloads.

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.