GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel libceph: osdmap_decode() under-counts osd_weight, allowing out-of-bounds reads

CVSS 9.1CVE-2026-52958Kernel, userspace & hypervisorcurated

Impact

osdmap_decode() decodes osd_state and osd_weight once per OSD (max_osd times) but its ceph_decode_need() check accounted for sizeof(*map->osd_weight) only once. A corrupted osdmap whose max_osd exceeds the real message contents therefore walks past the end of the buffer. This is the kernel Ceph client on the compute node parsing a map update from the monitors, so on a GPU fleet the consequence is a kernel-context out-of-bounds read - a node crash that takes down the jobs on it - triggered from the storage network rather than from tenant workloads. Nodes without CephFS or RBD mounts are unaffected.

Who can reach it

A compromised Ceph monitor or an attacker able to forge or modify osdmap updates on the storage network reaching a node's kernel Ceph client. No node credentials needed.

What to do

Update the kernel to a stable release where the decode check multiplies by map->max_osd (commits in the record), then drain and reboot each node with a Ceph mount. No runtime mitigation exists in the client; restrict the Ceph monitor network to trusted hosts and use msgr2 secure mode in the meantime.

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.