Executive brief
A vulnerability in the Linux kernel's NVMe storage driver could lead to a system deadlock, causing the operating system to become unresponsive. This issue occurs during specific storage management tasks, such as scanning disk partitions on systems using multiple paths to reach storage devices. If triggered, this could result in a complete loss of availability for the affected server, impacting business operations and data access.
Technical details
A deadlock condition was identified in the Linux kernel's NVMe multipath implementation (drivers/nvme/host/multipath.c) due to circular dependencies between disk->open_mutex, the kblockd workqueue, and partition_scan_work. The issue is triggered when partition scanning is deferred and executed via the generic kblockd workqueue, which can conflict with other block layer operations requiring the same mutex. The fix involves migrating partition_scan_work from the shared kblockd workqueue to the dedicated nvme_wq to break the dependency chain. This vulnerability primarily impacts system availability (Denial of Service). Patches have been released for stable kernel branches including 6.1.y, 6.6.y, and 6.12.y.
Affected products
- Linux Linux Kernel 6.1.118 to 6.1.159, 6.6.62 to 6.6.118, 6.11.9 to 6.12.0
Timeline
- 2025-11-17: patched: Initial fix committed to mainline kernel.
- 2025-12-16: disclosed: CVE published.
References
- https://git.kernel.org/stable/c/6d87cd5335784351280f82c47cc8a657271929c3
- https://git.kernel.org/stable/c/89456dab7ba5ab63d60945440926673a3205e829
- https://git.kernel.org/stable/c/b03eb63288a8ffe3adfb34e68309c8e2edb06d0b
- https://git.kernel.org/stable/c/e2a897ad5f538d314955c747a0a2edb184fcdecd
- https://git.kernel.org/stable/c/ef4ab2a8abe554379e10303ae86f7c501336ba0d