GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel NFS: backchannel request racing callback shutdown leaks transport refs or queues onto a freed svc_serv

UnscoredCVE-2026-89546Kernel, userspace & hypervisorcurated

Impact

A NFSv4 backchannel receive can complete while the callback service is being torn down. If bc_serv is already cleared, the committed request is neither enqueued nor freed and the transport reference taken on entry is leaked; if the producer wins the race, it can enqueue onto sv_cb_list after the callback threads have stopped, leaving a request linked to a svc_serv that is about to be freed. On a GPU node that unmounts NFSv4 while the server is still sending callbacks - a normal event during job teardown or a storage failover - this is either a slow leak or a use-after-free in the host kernel.

Who can reach it

Driven by the NFSv4 server: a backchannel request arriving as the client shuts its callback service down. Any server the node mounts, or a peer able to answer for it, can time traffic into that window; no separate client-side authentication is involved.

What to do

Install a stable kernel with xprt_svc_shutdown_bc() and the sv_cb_list drain (commits in the record). No runtime workaround. Rolling the kernel means draining and rebooting each NFSv4 client node.

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.