GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel SUNRPC: gssx decode error paths NULL-deref and leak group_info on the NFS server

CVSS 7.5CVE-2026-89544Kernel, userspace & hypervisorcurated

Impact

On an NFS server using Kerberos via gss-proxy, a failed allocation or a malformed gssx response leaves the decoded option array with count=1 and a NULL data pointer, and the caller dereferences data[0].value - a kernel oops that takes the file server down. The same error paths free a partially decoded svc_cred with a bare kfree, leaking the refcounted, kvmalloc-backed group_info on every failure. For a GPU fleet this matters where krb5 NFS backs home directories, datasets or checkpoints: the crash is on the server, so every node mounting it stalls at once, and recovery is a reboot of a machine that many jobs depend on.

Who can reach it

Reachable through the gss-proxy upcall path on an NFS server running krb5 security flavours; an unauthenticated client can drive context establishment, but the record does not show a clean remote trigger - the documented failure is an allocation failure or a bad gssx reply.

What to do

Update the host kernel to a stable release containing the fix (three stable branches carry it; see the git.kernel.org commits) and reboot the NFS server. No live mitigation is described beyond not running krb5 NFS; the fix is kernel-side only, so the server must be drained and rebooted.

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.