GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux nfsd: NFSD_A_SOCK_ADDR has no minimum length, giving a 12-byte out-of-bounds read in listener_set

UnscoredCVE-2026-89700Kernel, userspace & hypervisorcurated

Impact

nfsd_sock_nl_policy declared NFSD_A_SOCK_ADDR as bare NLA_BINARY with no minimum length, so a caller could send a 16-byte address with sa_family=AF_INET6 and cause a 12-byte out-of-bounds read across rpc_cmp_addr_port, svc_find_listener and kernel_bind. The listener configuration was also validated inline in two separate loops interleaved with mutating the running config, so a malformed entry was handled inconsistently and could leave the server's listeners half-changed. The gating capability is CAP_NET_ADMIN, which makes this a hardening fix rather than a tenant-reachable bug - but it matters on nodes where a container or automation holds CAP_NET_ADMIN in the host network namespace. The fix adds nfsd_nl_validate_listeners(), which walks the whole list before taking nfsd_mutex so a malformed request fails with no side effects.

Who can reach it

Local caller holding CAP_NET_ADMIN in the network namespace of the NFS server - root, or a privileged container with that capability. Not reachable by an ordinary tenant or by a remote NFS client.

What to do

Patch the kernel on in-kernel NFS server nodes and reboot. Low urgency given the CAP_NET_ADMIN requirement; fold it into the next scheduled kernel roll rather than opening a window for it. In the meantime, audit which workloads on the NFS server hold CAP_NET_ADMIN in the host netns.

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.