GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel ice: NULL dereference in the E810 low-latency PTP timestamp interrupt during reset

CVSS 7.8CVE-2025-39854Kernel, userspace & hypervisorcurated

Impact

ice_ll_ts_intr() touches the Tx timestamp tracker without checking that it is initialized, so a low-latency PTP timestamp interrupt racing the driver reset path gives a NULL dereference or use-after-free in interrupt context. Intel E810 is a common 100G NIC on GPU nodes, and PTP hardware timestamping is routinely enabled for training-cluster time sync, so the racing condition is a normal operational event (a NIC reset under load), not an attacker action. The consequence is a node panic - on a GPU host that means losing whatever job was resident and an unplanned drain. This is the companion of CVE-2025-39855, which fixes the same missing check in the standard timestamp IRQ handler; both were fixed separately because they are different handlers.

Who can reach it

Not a tenant-reachable path. Triggered by a race between an E810 low-latency Tx timestamp interrupt and the driver reset flow; requires the ice driver with PTP low-latency firmware support in use.

What to do

Update to a stable kernel carrying the fix (three backports linked in the record) and reboot the affected nodes; a driver-level fix on the interrupt path cannot be hot-applied. No fixed version string is given in the record - match the linked commits against the stable branch you run. Nodes without E810 NICs or with PTP disabled are not exposed.

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.