Re: [PATCH v3 sched_ext/for-7.3] sched_ext: Bound per-task reenqueues and eject the owning scheduler

[Andrea Righi <arighi@nvidia.com>]

Hi Tejun,

On Sun, Jul 26, 2026 at 09:49:54AM -1000, Tejun Heo wrote:
> Unlike local reenqueues, cap rejections have no repeat limit. A
> malfunctioning scheduler can keep re-inserting a task to a cid it lacks caps
> on, cycling the task through reject and reenqueue. This was assumed safe
> because a task that never runs trips the stall watchdog. However, the
> reenqueue irq_work re-arms itself and outranks the timer vector, blocking
> everything else on the CPU including stall detection and recovery, until the
> NMI hardlockup detector fires.
> 
> Local reenqueues already have a repeat cap, SCX_REENQ_LOCAL_MAX_REPEAT,
> which needs generalizing to cover all reenqueues. It also has an attribution
> problem. Counted per-cpu on root, it tears down the whole hierarchy even
> when a sub-scheduler caused the repeated reenqueues.
> 
> Generalize by bounding every reenqueue with one per-task counter. reenq_cnt
> is bumped in scx_do_enqueue_task() on each SCX_ENQ_REENQ, the single path
> every reenqueue producer passes through, and cleared in clr_task_runnable()
> when the task is picked to run and in scx_disable_task() when it leaves the
> scheduler's control. Past SCX_REENQ_MAX_REPEAT the task's owning scheduler
> is ejected with a new SCX_EXIT_ERROR_REENQ and the task is left stranded to
> be picked up during sched exit.
> 
> The SCX_EV_REENQ_LOCAL_REPEAT event becomes SCX_EV_REENQ_REPEAT, counting
> repeat reenqueues from all sources.
> 
> v2: Count SCX_EV_REENQ_REPEAT only when a reenqueue leads to another
>     reenqueue, not on every reenqueue.
> 
> v3: - Also clear reenq_cnt in scx_disable_task() so that the count doesn't
>       carry over to the next owner across sched class switches, scheduler
>       replacement or sub-scheduler rehoming (Andrea Righi).
> 
>     - Update the stale SCX_EV_REENQ_LOCAL_REPEAT references in sched-ext.rst
>       (Andrea Righi).
> 
> Signed-off-by: Tejun Heo <tj@kernel.org>

Ack about counting all reenqueues.

One minor nit: tools/sched_ext/include/scx/enum_defs.autogen.h still defines
HAVE_SCX_REENQ_LOCAL_MAX_REPEAT, should be renamed to HAVE_SCX_REENQ_MAX_REPEAT.

Other than that, looks good to me.

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

Thanks,
-Andrea

> ---
>  Documentation/scheduler/sched-ext.rst |    8 ++--
>  include/linux/sched/ext.h             |    1 
>  kernel/sched/ext/ext.c                |   58 +++++++++++++++++++---------------
>  kernel/sched/ext/internal.h           |   19 ++++-------
>  kernel/sched/ext/sub.c                |    6 +--
>  kernel/sched/ext/types.h              |    2 -
>  kernel/sched/sched.h                  |    1 
>  7 files changed, 51 insertions(+), 44 deletions(-)
> 
> --- a/Documentation/scheduler/sched-ext.rst
> +++ b/Documentation/scheduler/sched-ext.rst
> @@ -106,7 +106,7 @@ counters. Each counter occupies one ``na
>      SCX_EV_ENQ_SKIP_EXITING 0
>      SCX_EV_ENQ_SKIP_MIGRATION_DISABLED 0
>      SCX_EV_REENQ_IMMED 0
> -    SCX_EV_REENQ_LOCAL_REPEAT 0
> +    SCX_EV_REENQ_REPEAT 0
>      SCX_EV_REFILL_SLICE_DFL 456789
>      SCX_EV_BYPASS_DURATION 0
>      SCX_EV_BYPASS_DISPATCH 0
> @@ -129,9 +129,9 @@ The counters are described in ``kernel/s
>    ``SCX_OPS_ENQ_MIGRATION_DISABLED`` is not set).
>  * ``SCX_EV_REENQ_IMMED``: a task dispatched with ``SCX_ENQ_IMMED`` was
>    re-enqueued because the target CPU was not available for immediate execution.
> -* ``SCX_EV_REENQ_LOCAL_REPEAT``: a reenqueue of the local DSQ triggered
> -  another reenqueue; recurring counts indicate incorrect ``SCX_ENQ_REENQ``
> -  handling in the BPF scheduler.
> +* ``SCX_EV_REENQ_REPEAT``: a reenqueue led to another reenqueue without the
> +  task running in between; recurring counts indicate that the BPF scheduler
> +  keeps re-deciding placements it can't honor.
>  * ``SCX_EV_REFILL_SLICE_DFL``: a task's time slice was refilled with the
>    default value (``SCX_SLICE_DFL``).
>  * ``SCX_EV_BYPASS_DURATION``: total nanoseconds spent in bypass mode.
> --- a/include/linux/sched/ext.h
> +++ b/include/linux/sched/ext.h
> @@ -198,6 +198,7 @@ struct sched_ext_entity {
>  	u32			dsq_flags;	/* protected by DSQ lock */
>  	u32			flags;		/* protected by rq lock */
>  	u32			weight;
> +	u32			reenq_cnt;	/* reenqueues since last run */
>  	s32			sticky_cpu;
>  	s32			holding_cpu;
>  	s32			selected_cpu;
> --- a/kernel/sched/ext/ext.c
> +++ b/kernel/sched/ext/ext.c
> @@ -1904,6 +1904,24 @@ void scx_do_enqueue_task(struct rq *rq,
>  	p->scx.flags &= ~SCX_TASK_IMMED;
>  
>  	/*
> +	 * A task reenqueued too many times without running means the scheduler
> +	 * keeps re-deciding a placement it can't honor, e.g. re-inserting to a
> +	 * cid it lacks caps on. Eject the owning scheduler and strand the task
> +	 * to be picked up during sched exit.
> +	 */
> +	if (enq_flags & SCX_ENQ_REENQ) {
> +		if (++p->scx.reenq_cnt > 1)
> +			__scx_add_event(sch, SCX_EV_REENQ_REPEAT, 1);
> +
> +		if (unlikely(p->scx.reenq_cnt > SCX_REENQ_MAX_REPEAT)) {
> +			__scx_exit(sch, SCX_EXIT_ERROR_REENQ, 0, cpu_of(rq),
> +				   "%s[%d] reenqueued %u times without running",
> +				   p->comm, p->pid, p->scx.reenq_cnt);
> +			return;
> +		}
> +	}
> +
> +	/*
>  	 * If !scx_rq_online(), we already told the BPF scheduler that the CPU
>  	 * is offline and are just running the hotplug path. Don't bother the
>  	 * BPF scheduler.
> @@ -2025,8 +2043,10 @@ static void clr_task_runnable(struct tas
>  {
>  	list_del_init(&p->scx.runnable_node);
>  	WRITE_ONCE(p->scx.runnable_cpu, -1);
> -	if (reset_runnable_at)
> +	if (reset_runnable_at) {
>  		p->scx.flags |= SCX_TASK_RESET_RUNNABLE_AT;
> +		p->scx.reenq_cnt = 0;
> +	}
>  }
>  
>  static void enqueue_task_scx(struct rq *rq, struct task_struct *p, int core_enq_flags)
> @@ -3669,6 +3689,7 @@ static void scx_disable_task(struct scx_
>  	 */
>  	p->scx.dsq_vtime = 0;
>  	set_task_slice(p, 0);
> +	p->scx.reenq_cnt = 0;
>  
>  	/*
>  	 * Verify the task is not in BPF scheduler's custody. If flag
> @@ -4066,8 +4087,8 @@ static void process_ddsp_deferred_locals
>   * Reenqueued tasks go through ops.enqueue() with %SCX_ENQ_REENQ |
>   * %SCX_TASK_REENQ_IMMED. If the BPF scheduler dispatches back to the same local
>   * DSQ with %SCX_ENQ_IMMED while the CPU is still unavailable, this triggers
> - * another reenq cycle. Repetitions are bounded by %SCX_REENQ_LOCAL_MAX_REPEAT
> - * in process_deferred_reenq_locals().
> + * another reenq cycle. Repetitions are bounded by %SCX_REENQ_MAX_REPEAT in
> + * scx_do_enqueue_task(), which ejects the task's owning scheduler.
>   */
>  static bool local_task_should_reenq(struct rq *rq, struct task_struct *p,
>  				    u64 *reenq_flags, u32 *reason)
> @@ -4175,14 +4196,16 @@ static u32 reenq_local(struct scx_sched
>  
>  static void process_deferred_reenq_locals(struct rq *rq)
>  {
> -	u64 seq = ++rq->scx.deferred_reenq_locals_seq;
> -
>  	lockdep_assert_rq_held(rq);
>  
> +	/*
> +	 * A task can be re-queued within this loop when a reenqueued task
> +	 * bounces straight back to the local DSQ. That recursion is bounded by
> +	 * the per-task reenqueue cap in scx_do_enqueue_task().
> +	 */
>  	while (true) {
>  		struct scx_sched *sch;
>  		u64 reenq_flags;
> -		bool skip = false;
>  
>  		scoped_guard (raw_spinlock, &rq->scx.deferred_reenq_lock) {
>  			struct scx_deferred_reenq_local *drl =
> @@ -4201,27 +4224,12 @@ static void process_deferred_reenq_local
>  			reenq_flags = drl->flags;
>  			WRITE_ONCE(drl->flags, 0);
>  			list_del_init(&drl->node);
> -
> -			if (likely(drl->seq != seq)) {
> -				drl->seq = seq;
> -				drl->cnt = 0;
> -			} else {
> -				if (unlikely(++drl->cnt > SCX_REENQ_LOCAL_MAX_REPEAT)) {
> -					scx_error(sch, "SCX_ENQ_REENQ on SCX_DSQ_LOCAL repeated %u times",
> -						  drl->cnt);
> -					skip = true;
> -				}
> -
> -				__scx_add_event(sch, SCX_EV_REENQ_LOCAL_REPEAT, 1);
> -			}
>  		}
>  
> -		if (!skip) {
> -			/* see schedule_dsq_reenq() */
> -			smp_mb();
> +		/* see schedule_dsq_reenq() */
> +		smp_mb();
>  
> -			reenq_local(sch, rq, reenq_flags);
> -		}
> +		reenq_local(sch, rq, reenq_flags);
>  	}
>  }
>  
> @@ -5941,6 +5949,8 @@ static const char *scx_exit_reason(enum
>  		return "scx_bpf_error";
>  	case SCX_EXIT_ERROR_STALL:
>  		return "runnable task stall";
> +	case SCX_EXIT_ERROR_REENQ:
> +		return "reenqueue limit";
>  	default:
>  		return "<UNKNOWN>";
>  	}
> --- a/kernel/sched/ext/internal.h
> +++ b/kernel/sched/ext/internal.h
> @@ -56,6 +56,7 @@ enum scx_exit_kind {
>  	SCX_EXIT_ERROR = 1024,	/* runtime error, error msg contains details */
>  	SCX_EXIT_ERROR_BPF,	/* ERROR but triggered through scx_bpf_error() */
>  	SCX_EXIT_ERROR_STALL,	/* watchdog detected stalled runnable tasks */
> +	SCX_EXIT_ERROR_REENQ,	/* task hit reenqueue limit without running */
>  };
>  
>  /*
> @@ -1119,15 +1120,13 @@ struct scx_event_stats {
>  	s64		SCX_EV_REENQ_IMMED;
>  
>  	/*
> -	 * The number of times a reenq of local DSQ caused another reenq of
> -	 * local DSQ. This can happen when %SCX_ENQ_IMMED races against a higher
> -	 * priority class task even if the BPF scheduler always satisfies the
> -	 * prerequisites for %SCX_ENQ_IMMED at the time of enqueue. However,
> -	 * that scenario is very unlikely and this count going up regularly
> -	 * indicates that the BPF scheduler is handling %SCX_ENQ_REENQ
> -	 * incorrectly causing recursive reenqueues.
> +	 * The number of times a reenqueue (%SCX_ENQ_REENQ) led to another
> +	 * reenqueue without the task running in between. This count climbing
> +	 * rapidly indicates that the BPF scheduler keeps re-deciding placements
> +	 * it can't honor. A single task reenqueued more than
> +	 * %SCX_REENQ_MAX_REPEAT times gets its owning scheduler ejected.
>  	 */
> -	s64		SCX_EV_REENQ_LOCAL_REPEAT;
> +	s64		SCX_EV_REENQ_REPEAT;
>  
>  	/*
>  	 * Total number of times a task's time slice was refilled with the
> @@ -1221,7 +1220,7 @@ struct scx_event_stats {
>  	SCX_EVENT(SCX_EV_ENQ_SKIP_EXITING);				\
>  	SCX_EVENT(SCX_EV_ENQ_SKIP_MIGRATION_DISABLED);			\
>  	SCX_EVENT(SCX_EV_REENQ_IMMED);					\
> -	SCX_EVENT(SCX_EV_REENQ_LOCAL_REPEAT);				\
> +	SCX_EVENT(SCX_EV_REENQ_REPEAT);					\
>  	SCX_EVENT(SCX_EV_REFILL_SLICE_DFL);				\
>  	SCX_EVENT(SCX_EV_SLICE_CLAMPED);				\
>  	SCX_EVENT(SCX_EV_SLICE_DENIED);					\
> @@ -1260,8 +1259,6 @@ struct scx_dsp_ctx {
>  struct scx_deferred_reenq_local {
>  	struct list_head	node;
>  	u64			flags;
> -	u64			seq;
> -	u32			cnt;
>  };
>  
>  struct scx_sched_pcpu {
> --- a/kernel/sched/ext/sub.c
> +++ b/kernel/sched/ext/sub.c
> @@ -319,9 +319,9 @@ bool scx_task_reenq_on_cap_revoke(struct
>   * Drain @rq->scx.reject_dsq, reenqueueing each task so the BPF re-decides
>   * from p->scx.reenq_reason_*.
>   *
> - * A task can be re-rejected repeatedly, and there's no repeat limit here.
> - * Rejection can't happen for root, and sub-scheds can be safely ejected after
> - * triggering the stall watchdog.
> + * A task can be re-rejected repeatedly. The reenqueue is bounded per task in
> + * scx_do_enqueue_task(), which ejects the owning sub past SCX_REENQ_MAX_REPEAT.
> + * Rejection can't happen for root.
>   */
>  void scx_reenq_reject(struct rq *rq)
>  {
> --- a/kernel/sched/ext/types.h
> +++ b/kernel/sched/ext/types.h
> @@ -41,7 +41,7 @@ enum scx_consts {
>  	SCX_BYPASS_LB_MIN_DELTA_DIV	= 4,
>  	SCX_BYPASS_LB_BATCH		= 256,
>  
> -	SCX_REENQ_LOCAL_MAX_REPEAT	= 256,
> +	SCX_REENQ_MAX_REPEAT		= 256,
>  
>  	SCX_SUB_MAX_DEPTH		= 4,
>  };
> --- a/kernel/sched/sched.h
> +++ b/kernel/sched/sched.h
> @@ -823,7 +823,6 @@ struct scx_rq {
>  	struct list_head	sched_pcpus_to_kick;	/* see kick_cpus_irq_workfn() */
>  
>  	raw_spinlock_t		deferred_reenq_lock;
> -	u64			deferred_reenq_locals_seq;
>  	struct list_head	deferred_reenq_locals;	/* scheds requesting reenq of local DSQ */
>  	struct list_head	deferred_reenq_users;	/* user DSQs requesting reenq */
>  	struct balance_callback	deferred_bal_cb;