CVE-2026-64474 in Linux
Summary
by MITRE • 07/25/2026
In the Linux kernel, the following vulnerability has been resolved:
vfio: prevent infinite loop in vfio_mig_get_next_state() on blocked arc
vfio_mig_get_next_state() walks vfio_from_fsm_table[] one step at a time,
looping to skip optional states the device does not support until *next_fsm is supported. A blocked transition is encoded as VFIO_DEVICE_STATE_ERROR, which the trailing return reports as -EINVAL.
The skip loop does not account for the ERROR sentinel. state_flags_table[ERROR] is ~0U and vfio_from_fsm_table[ERROR][*] is
ERROR, so once *next_fsm becomes ERROR the loop condition stays true and *next_fsm never changes. The blocked arcs STOP_COPY -> PRE_COPY and STOP_COPY -> PRE_COPY_P2P map to ERROR yet pass the support check on a precopy-capable device, causing the loop to spin forever while holding the driver state mutex. This can result in a soft lockup, and a panic with softlockup_panic set.
Terminate the skip loop on the ERROR sentinel so a blocked transition falls through to the existing return and reports -EINVAL.
If you want to get the best quality for vulnerability data then you always have to consider VulDB.
Analysis
by VulDB Data Team • 07/27/2026
The vulnerability resides within the Linux kernel's vfio (Virtual Function I/O) subsystem, specifically in the device migration handling mechanism. This issue affects how the kernel manages state transitions during device migration operations, particularly when dealing with blocked transitions between different migration states. The problem manifests in the vfio_mig_get_next_state() function which is responsible for determining the next valid state in the device migration finite state machine. When a device attempts to transition through certain blocked paths, the function enters an infinite loop that can lead to system instability.
The technical flaw stems from how the function processes the vfio_from_fsm_table array, which contains mappings between current and next states for device migration. The implementation walks this table one step at a time, skipping over optional states that the device does not support by checking support status before proceeding. However, the skip loop logic fails to properly handle the ERROR sentinel value which represents blocked transitions. When a transition maps to VFIO_DEVICE_STATE_ERROR, typically encoded as an error condition, the function's loop condition remains true because state_flags_table[ERROR] equals ~0U and vfio_from_fsm_table[ERROR][*] also equals ERROR.
This creates a scenario where once *next_fsm becomes set to the ERROR value, the loop continues indefinitely without advancing the state pointer. The problematic transitions involve STOP_COPY -> PRE_COPY and STOP_COPY -> PRE_COPY_P2P mappings which are designed to be blocked but still pass support checks on devices that are capable of precopy operations. The loop operates while holding the driver state mutex, creating a critical race condition that can result in system lockups. According to common vulnerability classification standards, this represents a denial of service vulnerability with potential for system panic conditions.
The operational impact of this vulnerability is severe as it can cause a soft lockup of the entire system when the infinite loop occurs while holding critical kernel mutexes. Systems with softlockup_panic set configured will trigger immediate kernel panics, making this particularly dangerous in production environments. The issue affects any device utilizing VFIO migration functionality and can be triggered through normal device migration operations, particularly when attempting to transition between specific states that map to blocked arcs. This vulnerability directly impacts system availability and stability.
Mitigation strategies should focus on ensuring the skip loop properly terminates when encountering the ERROR sentinel value, allowing blocked transitions to fall through to the existing error reporting mechanism that returns -EINVAL. The fix requires modifying the loop condition logic in vfio_mig_get_next_state() to explicitly check for ERROR values and break out of the loop before they can cause indefinite spinning. This approach aligns with standard kernel security practices for preventing infinite loops in critical system paths. Organizations should prioritize updating their kernel versions to include this fix, particularly those running systems with VFIO-enabled devices that support migration features. The solution follows established patterns for handling sentinel values in finite state machine implementations and addresses the core issue identified in the Linux kernel's device migration subsystem.
This vulnerability demonstrates the importance of proper error condition handling in kernel-level finite state machines and highlights how seemingly minor logic flaws can lead to critical system instability. The fix ensures that blocked transitions are properly recognized and handled, preventing resource exhaustion that could otherwise lead to complete system unresponsiveness. The implementation should be tested thoroughly across different device types and migration scenarios to ensure the fix does not introduce regressions in normal operation while effectively resolving the infinite loop condition.