GPU VulnDB

Database/Control plane, storage & DevOps

Linux nfsd sunrpc: one failed group lookup strips a uid's supplementary groups for the cache TTL

UnscoredCVE-2026-99351Control plane, storage & DevOpscurated

Impact

When rpc.mountd cannot resolve a uid (for example while sssd or winbind is briefly unreachable), it returns zero groups, and the kernel caches that as a valid entry. For up to mountd's TTL (30 minutes by default), every request from that uid on every export runs with no supplementary groups, RPCSEC_GSS included, long after the directory service has recovered. On HPC and GPU clusters with NFS home or project directories under group permissions, users and jobs lose access to shared data. It fails closed: access is denied, not granted.

Who can reach it

No attacker needed. A transient NSS or directory-service failure on the NFS server triggers it for whoever is looked up during the outage.

What to do

Install a kernel with the stable fix (commits linked) on NFS servers and reboot them during a storage maintenance window. Until then, flushing the auth.unix.gid cache after a directory-service blip restores groups. The record does not give fixed versions.

References

Related entries

All Control plane, storage & DevOps 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.