GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel qla2xxx: re-initializing a queued work item on a repeated FC-NVMe abort corrupts the workqueue list

CVSS 8.8CVE-2026-89860Kernel, userspace & hypervisorcurated

Impact

The FC-NVMe abort paths called INIT_WORK() on priv->abort_work immediately before schedule_work(), which resets the work_struct list head and clears its pending bit. If a command is aborted twice - concurrent transport teardown plus a timeout-driven abort - the second INIT_WORK() reinitializes a work item that is still queued, corrupting the workqueue list and producing a crash or a looping worker. This is a race against error handling, so it surfaces exactly when the storage fabric is already unhappy: a node serving FC-NVMe-attached datasets panics or pins a CPU in a spinning worker during a storage event, and losing a GPU node mid-job is expensive. The fix moves the INIT_WORK() to command submission and leaves only schedule_work() in the abort paths.

Who can reach it

Not tenant-reachable. Needs FC-NVMe storage attached through a QLogic HBA and concurrent abort activity - typically a fabric fault, target reset or command timeout rather than a deliberate action. No authentication is involved.

What to do

Apply the linked stable commits or move to a patched distribution kernel and reboot each affected node. No configuration workaround exists; the only avoidance is not running FC-NVMe over qla2xxx.

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.