Executive brief
A flaw in the Linux kernel's RAID management system can cause a system deadlock when a RAID 4, 5, or 6 array is being resized or reshaped. If a user attempts to suspend the storage array while this process is active, the system may stop responding. This primarily affects system availability and could lead to a forced reboot.
Technical details
A deadlock vulnerability exists in the Linux kernel's Multiple Device (md) driver during RAID456 reshape operations. When direct I/O occurs across a reshape position, it may enter a sleep state in raid5_make_request() while holding an active_io reference. If the reshape is subsequently frozen and a suspend is initiated via md/suspend_lo or md/suspend_hi, the mddev_suspend() function waits for all in-flight I/O to drain. Because the reshape thread is frozen, the sleeping I/O never receives the progress it requires to wake up and drop its reference, resulting in a permanent hang. The fix ensures that reshape waiters are woken during the suspend sequence to allow them to abort and release their references.
Affected products
- Linux Linux Kernel 6.7 to 6.12.90, 6.18.32, 7.0.9
Timeline
- 2026-03-27: disclosed: Initial patch submission
- 2026-06-24: advisory: CVE published