In the Linux kernel, the following vulnerability has been resolved: sched_ext: Don't run ops.dequeue() with a DSQ lock held ops.dequeue() is invoked… (CVE-2026-98865)

The Linux kernel sched_ext component previously invoked ops.dequeue() while holding the source user DSQ lock on consume and move paths, leading to self-deadlocks in BPF schedulers that lock the source user DSQ during iteration (e.g., via bpf_iter_scx_dsq). The deadlock arises because all DSQ locks share one lockdep class, triggering recursion checks. The vulnerability was fixed by moving the ops.dequeue() invocation to after the DSQ unlock on all affected paths and clearing SCX_TASK_IN_CUSTODY under the lock to ensure a single callback invocation.

The vulnerability can cause self-deadlocks in the Linux kernel scheduler extension when a BPF scheduler locks the source user DSQ during ops.dequeue() invocation. This can lead to kernel hangs or degraded system responsiveness due to scheduler deadlock conditions. There is no indication of privilege escalation or data corruption from the provided data.

A fix has been implemented in the Linux kernel to move the ops.dequeue() call after the DSQ unlock on all affected paths, preventing the deadlock condition. Users should update to the patched kernel version once available. Patch status is not explicitly confirmed in the provided data; check the vendor advisory or Linux kernel mailing lists for the official fix and updated versions.