Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-97925 Overview
CVE-2026-97925 is a race condition in the Linux kernel's tick broadcast subsystem. The flaw occurs when a broadcast clockevent device is replaced concurrently with a broadcast tick operation on another CPU. The function tick_install_broadcast_device() is not serialized against ongoing tick broadcast operations, allowing one CPU to arm a clockevent that another CPU has already shut down and detached.
When the original broadcast device has a restricted interrupt affinity mask and the last CPU in that mask goes offline, the BUG() in tick_cleanup_dead_cpu() triggers because the clockevent device is not in a detached state. The issue has been resolved upstream through serialization using tick_broadcast_lock.
Critical Impact
Concurrent broadcast device replacement can trigger a kernel BUG() leading to denial of service when CPU hotplug and clockevent replacement race on systems with affinity-restricted broadcast timers.
Affected Products
- Linux kernel (upstream mainline and stable branches)
- Distributions shipping affected tick/broadcast clockevent code paths
- Systems using broadcast clockevent devices with restricted CPU affinity masks
Discovery Timeline
- 2026-09-25 - CVE-2026-97925 published to NVD
- 2026-09-25 - Last updated in NVD database
Technical Details for CVE-2026-97925
Vulnerability Analysis
The vulnerability resides in the Linux kernel timer tick broadcast infrastructure. The tick broadcast device wakes CPUs from deep idle states when their per-CPU timers are stopped. When the broadcast device is replaced through tick_install_broadcast_device(), the operation calls clockevents_exchange_device() which shuts down the current device, detaches it, and installs the replacement.
The reporter 朱恺乾 identified that __tick_broadcast_oneshot_control() on CPUA can read tick_broadcast_device.evtdev into a local variable bc, while CPUB concurrently executes the exchange. After CPUB finishes shutdown and detach, CPUA proceeds to call tick_broadcast_set_event(bc, next_event) on the now-detached device. This arms a clockevent that is no longer in a valid operational state.
Root Cause
The root cause is missing serialization between tick_install_broadcast_device() and tick broadcast operations. The former did not acquire tick_broadcast_lock, leaving a window where the broadcast device pointer and its operational state could diverge across CPUs. This is a classic race condition [CWE-362] in kernel timekeeping code.
The fix splits clockevents_exchange_device() into two phases for the broadcast path. The exchange, shutdown, and detach happen under tick_broadcast_lock, while the module reference count is dropped after releasing the lock. Dropping the module reference under the lock is not permitted, which required the two-step restructuring.
Attack Vector
Exploitation requires local kernel conditions where a broadcast clockevent device is replaced while another CPU performs broadcast tick processing. Triggering the follow-on BUG() additionally requires the broadcast device to have a restricted interrupt affinity mask and the last CPU in that mask to go offline. The condition manifests as a kernel panic or forced reboot. See the upstream fix commits referenced below for the authoritative technical detail.
Detection Methods for CVE-2026-97925
Indicators of Compromise
- Kernel BUG() splats originating in tick_cleanup_dead_cpu() referencing a clockevent device not in CLOCK_EVT_STATE_DETACHED
- Unexpected kernel panics correlated with CPU hotplug (cpu offline) events on systems using broadcast timers with affinity restrictions
- Call traces including __tick_broadcast_oneshot_control, tick_broadcast_set_event, and clockevents_exchange_device
Detection Strategies
- Inventory running kernels against the fixed stable branches referenced in the upstream commits
- Review crash dumps and kdump artifacts for panics matching the tick broadcast call path
- Correlate CPU hotplug audit events with timer subsystem errors in dmesg and journal logs
Monitoring Recommendations
- Forward kernel logs to a centralized logging pipeline and alert on BUG: or kernel panic entries referencing tick_ or clockevents_ symbols
- Track CPU hotplug operations on virtualization hosts and bare-metal systems with constrained timer affinities
- Monitor unexpected host reboots on fleets that frequently exercise CPU online/offline transitions
How to Mitigate CVE-2026-97925
Immediate Actions Required
- Apply the upstream kernel patches referenced in the stable commits as soon as vendor builds are available
- Prioritize patching on systems that exercise CPU hotplug or run on hardware with affinity-restricted broadcast timers
- Validate that distribution kernel updates include the tick/broadcast: Plug clockevents replacement race fix
Patch Information
The upstream fix is distributed across the following stable branch commits: Kernel Git Commit 113a9796, Kernel Git Commit 292b7a8e, Kernel Git Commit 3525b94b, and Kernel Git Commit 3da55616. The fix serializes tick_install_broadcast_device() with tick_broadcast_lock and defers the module reference drop until after the lock is released.
Workarounds
- Avoid dynamic CPU hotplug operations on affected kernels where broadcast timers have restricted affinity masks
- Where feasible, use broadcast clockevent devices with CPU-wide affinity to reduce the probability of hitting the dead-CPU cleanup path
- Schedule maintenance windows for kernel updates rather than relying on long-term workarounds, as no complete mitigation exists outside the patch
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

