In the Linux kernel, the following vulnerability has been resolved: blk-cgroup: save IRQ state in blkg_tryget_closest() bio_set_dev() associates the… (CVE-2026-98832)

The Linux kernel blk-cgroup vulnerability (CVE-2026-98832) stems from improper interrupt state management in blkg_tryget_closest(). Specifically, bio_set_dev() associates a bio with a block group (blkg), and if the blkg lookup misses, blkg_tryget_closest() acquires the queue lock using spin_lock_irq() and releases it with spin_unlock_irq(), which unconditionally enables local interrupts. If the caller had interrupts disabled (e.g., dm-thin's pool_map()), this leads to interrupts being enabled while holding a spinlock, allowing an I/O completion softirq to run on the same CPU and re-acquire the same lock, causing a deadlock. Previous fixes were reverted due to changes in locking location. The correct fix involves saving and restoring the caller's IRQ state to avoid enabling interrupts unexpectedly.

This vulnerability can cause deadlocks in the Linux kernel blk-cgroup subsystem when interrupts are enabled while holding a spinlock, leading to inconsistent lock usage and potential system hangs or degraded performance. It affects kernel stability and reliability but does not directly indicate privilege escalation or data corruption.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. The fix involves saving and restoring the caller's IRQ state in blkg_tryget_closest() to prevent enabling interrupts while holding locks. Users should monitor official Linux kernel advisories for patches addressing this issue.