Re: [PATCH] sched/core: initialize resched_latency in sched_tick()

[Steven Rostedt]

On Tue, 17 Mar 2026 15:58:07 -0400
Joseph Salisbury <joseph.salisbury@xxxxxxxxxx> wrote:

> sched_tick() only assigns resched_latency when the LATENCY_WARN feature is
> enabled, but tests the variable later in a separate feature check.
> 
> That leaves the code dependent on both checks observing identical state.
> Initializing resched_latency to zero makes the later test safe even if the
> feature state changes between the two checks.
> 
> This does not change behavior when LATENCY_WARN is enabled. It only avoids
> using an uninitialized local value.
> 
> Fixes: c006fac556e4 ("sched: Warn on long periods of pending need_resched")
> Cc: stable@xxxxxxxxxxxxxxx
> Assisted-by: Codex:GPT-5
> Signed-off-by: Joseph Salisbury <joseph.salisbury@xxxxxxxxxx>
> ---
>  kernel/sched/core.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 08a06162dd9a..15916186e4ff 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -5548,7 +5548,7 @@ void sched_tick(void)
>  	struct task_struct *donor;
>  	struct rq_flags rf;
>  	unsigned long hw_pressure;
> -	u64 resched_latency;
> +	u64 resched_latency = 0;
>  
>  	if (housekeeping_cpu(cpu, HK_TYPE_KERNEL_NOISE))
>  		arch_scale_freq_tick();

The code has:

	if (sched_feat(LATENCY_WARN))
		resched_latency = cpu_resched_latency(rq);
	calc_global_load_tick(rq);
	sched_core_tick(rq);
	scx_tick(rq);

	rq_unlock(rq, &rf);

	if (sched_feat(LATENCY_WARN) && resched_latency)
		resched_latency_warn(cpu, resched_latency);


I guess if LATENCY_WARN gets enabled between the first check and the
second, then the resched_latency could be uninitialized.

Heck, probably could even get rid of the second sched_feat(LATENCY_WARN),
and make it simply:

	if (resched_latency)

Reviewed-by: Steven Rostedt (Google) <rostedt@xxxxxxxxxxx>

-- Steve