GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel ice driver: double free of transmit skb on the TSO/checksum error path

CVSS 7.8CVE-2026-53009Kernel, userspace & hypervisorcurated

Impact

When ice_tso() or ice_tx_csum() fails, ice_xmit_frame_ring() frees the skb but leaves the tx_buf still marked ICE_TX_BUF_SKB, so bringing the interface down afterwards makes ice_clean_tx_ring() free the same skb a second time. Intel E810 NICs are common as the host/management and storage NIC on GPU nodes alongside the RDMA fabric adapters, so this sits in a driver that is loaded on production nodes. The practical trigger is narrow - the reporter needed injected failures under KASAN, and a following transmit normally overwrites the stale tx_buf - but a double free in the kernel allocator is a local memory-corruption primitive and interface-down is a routine operation during drain. There is no remote or fabric-side path here.

Who can reach it

Local, and in practice requires the ability to make the transmit offload path fail and then bring the interface down; the record describes reproduction only with injected failures. No remote attacker path.

What to do

Take the stable kernel fix (referenced commits) or the Red Hat errata kernels RHSA-2026:42919 / 54246 / 54247, then reboot each node running the ice driver. Nodes can be patched during the normal drain-and-reboot rotation; nothing in the record suggests urgency ahead of that.

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.