CVE-2026-90003 PUBLISHED

futex: Prevent rcuwait use-after-free during requeue PI

Assigner: Linux
Reserved: 11.09.2026 Published: 16.09.2026 Updated: 16.09.2026

In the Linux kernel, the following vulnerability has been resolved:

futex: Prevent rcuwait use-after-free during requeue PI

On PREEMPT_RT, FUTEX_CMP_REQUEUE_PI can trigger a KASAN report (slab-out-of-bounds) in futex_requeue_pi_complete() invocation of rcuwait_wake_up().

The futex_q used by futex_wait_requeue_pi() is allocated on the waiter's stack. An early wakeup can race with a PI requeue as follows:

<pre> waiter requeue task ------ ------------ </pre>

futex_wait_requeue_pi() futex_do_wait() schedule() futex_requeue futex_proxy_trylock_atomic() futex_requeue_pi_prepare() Q_REQUEUE_PI_NONE -> Q_REQUEUE_PI_IN_PROGRESS * timeout/ signal wakes waiter * futex_requeue_pi_wakeup_sync() Q_REQUEUE_PI_IN_PROGRESS -> Q_REQUEUE_PI_WAIT requeue_pi_wake_futex futex_requeue_pi_complete() cmpxchg Q_REQUEUE_PI_WAIT -> Q_REQUEUE_PI_LOCKED rcuwait_wait_event() if (atomic_read(&q->requeue_state) != Q_REQUEUE_PI_WAIT) break / no schedule() /

/ q.pi_state->owner == current / futex_private_hash_put() / return from syscall / rcuwait_wake_up(&q->requeue_wait) / q is gone /

futex_requeue_pi_complete() publishes Q_REQUEUE_PI_LOCKED before calling rcuwait_wake_up(). The waiter observes this state in rcuwait_wait_event() before invoking schedule() in rcuwait_wait_event(). Here, the waiter is free leave the syscall before requeue task can complete the wake.

To address this race skip rcuwait_wake_up() in the Q_REQUEUE_PI_LOCKED case. This state is only published by requeue_pi_wake_futex(), which saves q->task before futex_requeue_pi_complete() and wakes the waiter via wake_up_state().

This wake is intended to wake the waiter from its futex_do_wait() sleep. If the waiter is still sleeping there, it can not get into the Q_REQUEUE_PI_WAIT state (and require this removed wake). Should the waiter be woken up from futex_do_wait() by other means (as in this example) and sleep in futex_requeue_pi_wakeup_sync() then the wake_up_state() from requeue_pi_wake_futex() will wake it, too. Should the waiter task terminate before wake_up_state() had a chance to wake the task then the task pointer does not become invalid because the futex_hash_bucket::lock is held and the task pointer is RCU protected.

[bigeasy: Updated comment and commit message]

Metrics

CVSS Vector: CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
CVSS Score: 7.8

AV:L - The bug is reached only through the local futex(2) syscalls FUTEX_WAIT_REQUEUE_PI and FUTEX_CMP_REQUEUE_PI into kernel/futex/requeue.c; there is no network, Bluetooth, or physical-device path. AC:L - The attacker controls both sides of the race: a waiter thread using FUTEX_WAIT_REQUEUE_PI with an attacker-chosen timeout or signal, and a concurrent FUTEX_CMP_REQUEUE_PI from another thread, and can retry with CPU pinning until the waiter returns before rcuwait_wake_up() on PREEMPT_RT. PR:L - Any unprivileged local user can invoke futex() with FUTEX_WAIT_REQUEUE_PI and FUTEX_CMP_REQUEUE_PI; do_futex() and futex_requeue() perform no capability or credential checks, CONFIG_FUTEX_PI is default-y, and the path is reachable from containers without elevated privileges. UI:N - The attacker's own threads perform the wait, timeout or signal, and PI requeue; no victim action such as mounting a filesystem or opening a file is required. S:U - The use-after-free corrupts kernel memory (the waiter's stack-allocated futex_q and a type-confused task_struct) within the same kernel security authority; this is a standard local privilege-escalation issue, not a VM, IOMMU, or hypervisor escape. C:H - This is a use-after-free of the waiter's stack-allocated futex_q.requeue_wait; rcuwait_wake_up() reads a stale, attacker-refilled task_struct pointer and dereferences it in try_to_wake_up(), enabling arbitrary kernel-memory disclosure. I:H - The stale futex_q lives on the attacker's own kernel stack, which a following syscall can refill with controlled bytes, so rcuwait_wake_up() passes an attacker-chosen pointer into try_to_wake_up(), which writes task state and makes indirect scheduler calls, enabling privilege escalation. A:H - In the unexploited case, rcuwait_wake_up() dereferences a wild pointer from reused stack memory and causes a kernel oops or panic; any use-after-free of this kind is a High availability impact.

Product Status

Vendor Linux
Product Linux
Versions Default: unaffected
  • affected from 07d91ef510fb16a2e0ca7453222105835b7ba3b8 to 220ee9e04ca3b7f014c000264aa6c884f036c86e (excl.)
  • affected from 07d91ef510fb16a2e0ca7453222105835b7ba3b8 to 81aadbd09bf1dcd3238212f336ba699503557ae8 (excl.)
  • affected from 07d91ef510fb16a2e0ca7453222105835b7ba3b8 to 244f301759fd34b1dd0b4192ce44f8ef224e027d (excl.)
  • affected from 07d91ef510fb16a2e0ca7453222105835b7ba3b8 to 7d1559126d86be6e4f6a85663dfbfe85caa47e37 (excl.)
  • affected from 07d91ef510fb16a2e0ca7453222105835b7ba3b8 to e4a4ccfa470f910b747b3ee8d18670ed8ac8a236 (excl.)
  • affected from 07d91ef510fb16a2e0ca7453222105835b7ba3b8 to a3b8d46fe401cba3a5c46dea610e6eb3dc15370e (excl.)
Vendor Linux
Product Linux
Versions Default: affected
  • Version 5.15 is affected
  • unaffected from 0 to 5.15 (excl.)
  • unaffected from 6.1.188 to 6.1.* (incl.)
  • unaffected from 6.6.157 to 6.6.* (incl.)
  • unaffected from 6.12.110 to 6.12.* (incl.)
  • unaffected from 6.18.51 to 6.18.* (incl.)
  • unaffected from 7.2.5 to 7.2.* (incl.)
  • unaffected from 7.3-rc2 to * (incl.)

References