Re: [PATCH v2] sched/core: Fix a TOCTOU bug in sched_non_preferred_cpu_push_stop()
From: Peter Zijlstra
Date: Thu Nov 19 2026 - 04:07:56 EST
On Wed, Nov 18, 2026 at 02:59:21PM +0800, Fushuai Wang wrote:
> From: Fushuai Wang <wangfushuai@xxxxxxxxx>
>
> sched_push_current_non_preferred_cpu() is called in irq context. It
> checks is_migration_disabled() before queuing the stopper.
>
> migrate_disable() sets p->migration_disabled in preempt_disable() section.
> An irq can hit after preempt_disable() but before p->migration_disabled=1.
> After the irq returns, the task cannot be preempted by stopper becuase
> preemption is disabled, so it finishes setting p->migration_disabled=1
> before the stopper runs.
>
> It may be like:
> task context: irq context:
> migrate_disable()
> preempt_disable()
> is_migration_disabled()
> migration_disabled = 1
> preempt_enable()
>
> So recheck is_migration_disabled() in sched_non_preferred_cpu_push_stop()
> before calling __migrate_task().
>
*sigh*:
https://lore.kernel.org/r/20260925074953.GG4121339@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> Fixes: 74699f56ebcf ("sched/core: Push current task from non preferred CPU")
> Signed-off-by: Fushuai Wang <wangfushuai@xxxxxxxxx>
> ---
> v1->v2: No code changes, just improve the commit message.
>
> kernel/sched/core.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 5ed660f099a5..6f9374aa3f19 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -11314,7 +11314,8 @@ static int sched_non_preferred_cpu_push_stop(void *arg)
> rq->npc_push_work_pending = false;
> update_rq_clock(rq);
>
> - if (task_rq(p) == rq && task_on_rq_queued(p)) {
> + if (task_rq(p) == rq && task_on_rq_queued(p) &&
> + !is_migration_disabled(p)) {
> struct rq *dest_rq = __migrate_task(rq, &rf, p, cpu);
>
> if (rq != dest_rq)
I think I like a comment with this. Perhaps add something like:
/*
* Racy: is_migration_disabled(current) isn't stable.
*/
if (is_migration_disabled(push_task))
return;
in sched_push_current_non_preferred_cpu().
Also, I wonder if you'd find more of this if you stick something like:
lockdep_assert(!task_on_cpu(p) || in_task());
inside is_migration_disabled().