schedule_deferred_locked() skips scheduling a deferred action while
SCX_RQ_IN_WAKEUP is set and relies on the task_woken_scx() call that follows
a wakeup enqueue to run it. enqueue_task_scx() sets the flag from the merged
enqueue flags, which include the flags stashed for a remote activation.
move_remote_task_to_local_dsq() thus sets SCX_RQ_IN_WAKEUP on the
destination rq when the moved task was woken up, although no
task_woken_scx() follows that activation.

An IMMED insert into a busy destination requests a local reenqueue during
that enqueue. The request gets linked but not scheduled and stays pending
until an unrelated wakeup or preemption on that CPU runs the deferred
actions. The IMMED task sits behind the running task in the meantime. If
nothing runs them before the scheduler is disabled, the request outlives the
scheduler and points into its freed per-cpu area, which the next scheduler
dereferences from run_deferred().

Test the core enqueue flags for the wakeup bit. Only the core's wakeup path
is followed by task_woken_scx().

Fixes: 57ccf5ccdc56 ("sched_ext: Fix enqueue_task_scx() truncation of upper enqueue flags")
Cc: stable@vger.kernel.org # v7.1+
Reported-by: Andrea Righi <arighi@nvidia.com>
Link: https://lore.kernel.org/all/20260916145807.3250167-1-arighi@nvidia.com/
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 kernel/sched/ext/ext.c |    7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

--- a/kernel/sched/ext/ext.c
+++ b/kernel/sched/ext/ext.c
@@ -2141,7 +2141,12 @@ static void enqueue_task_scx(struct rq *
 	int sticky_cpu = p->scx.sticky_cpu;
 	u64 enq_flags = core_enq_flags | rq->scx.remote_activate_enq_flags;
 
-	if (enq_flags & ENQUEUE_WAKEUP)
+	/*
+	 * SCX_RQ_IN_WAKEUP promises a task_woken_scx() call once this enqueue
+	 * returns. Only the core's wakeup path delivers one. The flags stashed
+	 * for a remote activation may carry the wakeup bit without it.
+	 */
+	if (core_enq_flags & ENQUEUE_WAKEUP)
 		rq->scx.flags |= SCX_RQ_IN_WAKEUP;
 
 	/*

Applied to sched_ext/for-7.3-fixes.

Thanks.

--
tejun

Hi Tejun,

On Wed, Sep 16, 2026 at 12:00:24PM -1000, Tejun Heo wrote:
> schedule_deferred_locked() skips scheduling a deferred action while
> SCX_RQ_IN_WAKEUP is set and relies on the task_woken_scx() call that follows
> a wakeup enqueue to run it. enqueue_task_scx() sets the flag from the merged
> enqueue flags, which include the flags stashed for a remote activation.
> move_remote_task_to_local_dsq() thus sets SCX_RQ_IN_WAKEUP on the
> destination rq when the moved task was woken up, although no
> task_woken_scx() follows that activation.
> 
> An IMMED insert into a busy destination requests a local reenqueue during
> that enqueue. The request gets linked but not scheduled and stays pending
> until an unrelated wakeup or preemption on that CPU runs the deferred
> actions. The IMMED task sits behind the running task in the meantime. If
> nothing runs them before the scheduler is disabled, the request outlives the
> scheduler and points into its freed per-cpu area, which the next scheduler
> dereferences from run_deferred().
> 
> Test the core enqueue flags for the wakeup bit. Only the core's wakeup path
> is followed by task_woken_scx().
> 
> Fixes: 57ccf5ccdc56 ("sched_ext: Fix enqueue_task_scx() truncation of upper enqueue flags")
> Cc: stable@vger.kernel.org # v7.1+
> Reported-by: Andrea Righi <arighi@nvidia.com>
> Link: https://lore.kernel.org/all/20260916145807.3250167-1-arighi@nvidia.com/
> Signed-off-by: Tejun Heo <tj@kernel.org>

This looks better than my fix.

Reviewed-by: Andrea Righi <arighi@nvidia.com>

Thanks,
-Andrea

> ---
>  kernel/sched/ext/ext.c |    7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> --- a/kernel/sched/ext/ext.c
> +++ b/kernel/sched/ext/ext.c
> @@ -2141,7 +2141,12 @@ static void enqueue_task_scx(struct rq *
>  	int sticky_cpu = p->scx.sticky_cpu;
>  	u64 enq_flags = core_enq_flags | rq->scx.remote_activate_enq_flags;
>  
> -	if (enq_flags & ENQUEUE_WAKEUP)
> +	/*
> +	 * SCX_RQ_IN_WAKEUP promises a task_woken_scx() call once this enqueue
> +	 * returns. Only the core's wakeup path delivers one. The flags stashed
> +	 * for a remote activation may carry the wakeup bit without it.
> +	 */
> +	if (core_enq_flags & ENQUEUE_WAKEUP)
>  		rq->scx.flags |= SCX_RQ_IN_WAKEUP;
>  
>  	/*