GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel libceph: unchecked OSD index from network packets panics the client node

CVSS 7.5CVE-2025-68283Kernel, userspace & hypervisorcurated

Impact

OSD indexes arriving in Ceph map messages were trusted without validation and reached a BUG_ON, so a malformed or hostile map crashes the kernel on any node mounting CephFS or RBD through the in-kernel client. On a GPU fleet that keeps datasets and checkpoints on Ceph, every compute node is a kernel Ceph client, so the blast radius is the nodes themselves and not the storage cluster. A kernel panic takes down whatever training or inference job held the node, and recovery is a reboot rather than a service restart. The fix replaces the BUG_ON with a bounds check against map->max_osd.

Who can reach it

Anyone able to put packets on the storage network that the kernel Ceph client accepts as map data - a compromised or spoofed monitor/OSD on the storage VLAN. No authentication to the node itself is needed; reachability of the storage path is the requirement.

What to do

Take the stable kernel containing the libceph bounds-check commit from your distribution and reboot each affected node; the fix is in kernel code loaded at boot, so there is no live-patch-free path short of a vendor hot-patch. Until then, keep the Ceph public network off any segment tenants can reach. No fixed mainline version is stated in the record beyond the listed stable commits.

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.