GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel CephFS client: leaked path buffers on error paths exhaust kernel memory

CVSS 5.5CVE-2026-43419Kernel, userspace & hypervisorcurated

Impact

ceph_mdsc_build_path() takes a name buffer from __getname() and, on several error paths, returned without calling __putname(), leaking the buffer whenever ownership was not handed to the caller via path_info.path. Each leak is one kernel name buffer; repeated under a workload that keeps hitting those error paths, it is unbounded kernel memory growth on any host with a CephFS mount. On GPU nodes that mount CephFS for datasets or checkpoints this is an availability problem on the node itself, and the memory is not reclaimable without a reboot. The record does not establish how easily the error paths are reached from userspace, so treat the rate as unknown.

Who can reach it

Local, through filesystem operations on a mounted CephFS. No special privilege beyond access to the mount is implied by the record; the host must be running the in-kernel CephFS client (the userspace/FUSE client and non-Ceph nodes are unaffected).

What to do

Take the stable kernel fix adding the missing __putname() calls (commits linked from the record). Rollout is a kernel upgrade - drain the node and reboot. Leaked memory on a running node is not recoverable short of a reboot. The record names no fixed release version. Nodes that do not mount CephFS in the kernel need no action.

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.