Re: [PATCH] sched/core: Convert remaining BUG_ON() instances to WARN_ON_ONCE()

[Peter Zijlstra <peterz@infradead.org>]

On Thu, Aug 27, 2026 at 01:40:14PM +0200, Amin Gattout wrote:
> Commit 09348d75a6ce ("sched/all: Change all BUG_ON() instances in the
> scheduler to WARN_ON_ONCE()") left a few BUG_ON() behind in core.c.
> Convert them as well.
> 
> As described in Documentation/process/deprecated.rst, BUG_ON() crashes
> the system, which makes the failure harder to report and to debug, as
> the message often never reaches the console or the syslog. With
> WARN_ON_ONCE() the user gets a chance to see and report the problem
> instead, and system owners who do not want to keep running after an
> "impossible" condition can still set panic_on_warn.
> 
> Signed-off-by: Amin Gattout <amin.gattout@gmail.com>
> ---
> Follow-up to commit 09348d75a6ce which converted the scheduler BUG_ON()
> instances to WARN_ON_ONCE() but left a few behind in core.c.
> 
> Build tested on arm64 with CONFIG_NO_HZ_FULL and CONFIG_SCHED_CLASS_EXT
> enabled so that every converted site is compiled.
> ---
>  kernel/sched/core.c | 20 ++++++++++----------
>  1 file changed, 10 insertions(+), 10 deletions(-)
> 
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index f78275192036..bab74b4f0dd6 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -5949,7 +5949,7 @@ static void sched_tick_stop(int cpu)
>  int __init sched_tick_offload_init(void)
>  {
>  	tick_work_cpu = alloc_percpu(struct tick_work);
> -	BUG_ON(!tick_work_cpu);
> +	WARN_ON_ONCE(!tick_work_cpu);
>  	return 0;
>  }
>  

> @@ -8954,13 +8954,13 @@ void __init sched_init(void)
>  	int i;
>  
>  	/* Make sure the linker didn't screw up */
> -	BUG_ON(!sched_class_above(&stop_sched_class, &dl_sched_class));
> -	BUG_ON(!sched_class_above(&dl_sched_class, &rt_sched_class));
> -	BUG_ON(!sched_class_above(&rt_sched_class, &fair_sched_class));
> -	BUG_ON(!sched_class_above(&fair_sched_class, &idle_sched_class));
> +	WARN_ON_ONCE(!sched_class_above(&stop_sched_class, &dl_sched_class));
> +	WARN_ON_ONCE(!sched_class_above(&dl_sched_class, &rt_sched_class));
> +	WARN_ON_ONCE(!sched_class_above(&rt_sched_class, &fair_sched_class));
> +	WARN_ON_ONCE(!sched_class_above(&fair_sched_class, &idle_sched_class));
>  #ifdef CONFIG_SCHED_CLASS_EXT
> -	BUG_ON(!sched_class_above(&fair_sched_class, &ext_sched_class));
> -	BUG_ON(!sched_class_above(&ext_sched_class, &idle_sched_class));
> +	WARN_ON_ONCE(!sched_class_above(&fair_sched_class, &ext_sched_class));
> +	WARN_ON_ONCE(!sched_class_above(&ext_sched_class, &idle_sched_class));
>  #endif
>  

These all really should be BUG_ON(), there is absolutely no point in
tying to complete the boot if they fail.