Linux kernel md/raid10 deadlock with check and nowait requests
CVE-2026-46050Medium
When a RAID10 array check operation is running, nowait requests bypass the normal barrier mechanism and return immediately with an error without incrementing the nr_pending counter. However, the kernel then calls allow_barrier() which decrements nr_pending unconditionally, causing the counter to underflow. This causes the md resync thread to deadlock whilst waiting for nr_pending to reach zero, blocking all I/O on the array. An attacker with local access to trigger both a check operation and nowait I/O requests on the same RAID10 array could cause a denial of service.
Plain English summary of the NVD and CISA text, checked by machine against both.
Scores and dates
CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
Detail
Affected versions
- linux linux kernel: from 5.15.189, before 5.15.209
- linux linux kernel: from 6.1.146, before 6.1.175
- linux linux kernel: from 6.6.99, before 6.6.140
- linux linux kernel: from 6.12.39, before 6.12.86
- linux linux kernel: from 6.15.7, before 6.16
- linux linux kernel: from 6.16.1, before 6.18.27
- linux linux kernel: from 6.19, before 7.0.4
- linux linux kernel: 6.16
Reports naming this CVE1 in total
· All CISA Advisories
Siemens SIMATIC S7-1500 TM MFP (opens in a new tab)