Technical impact
The flaw lies in the Linux kernel’s SCSI target subsystem: target_do_delayed_work() acquires the delayed_cmd_lock with a plain spin_lock while leaving interrupts enabled. When a command completes from an interrupt or backend context under a fabric that supports direct completion (complete_type=1) and the initiator sends an ORDERED task attribute, the same lock is later required by a work item queued on system_percpu_wq. Because the lock remains held with interrupts on, the work item spins indefinitely, causing a self‑deadlock that halts the kernel and results in a denial of service. The CVE specifies that the vulnerability can lead to a kernel hang, effectively taking down the affected system.
Affected Systems
This issue affects Linux kernel builds that include the SCSI target subsystem with vhost‑scsi or similar fabrics configured for direct completion. Systems running the default kernel with direct completion disabled or which never use the ORDERED task attribute are not affected; the problem manifests only when a fabric accepts ORDERED commands and processes completions from interrupt context. Virtual SCSI guests using virtio_scsi and vhost‑scsi that enable direct completion are especially vulnerable.
Risk and Exploitability
The CVSS score of 5.6 represents a medium severity because the flaw requires a specialized configuration. The EPSS score of 0.00145 (less than 1%) indicates a very low probability of exploitation in the wild, and the vulnerability is not listed in the CISA KEV catalog, suggesting no known active exploitation. However, if an attacker can issue ORDERED SCSI commands against a target that uses direct completion, they can trigger the lock‑deadlock and cause a kernel crash. Based on the description, it is inferred that the attacker must have the ability to send SCSI commands to the target from a client that can understand and transmit the ORDERED attribute; therefore the threat is limited to trusted or compromised initiators in environments exposing the vulnerable fabric.