Re: [PATCH v2 1/2] hung_task: Introduce touch_hung_task_detector().

[Masami Hiramatsu (Google) <mhiramat@kernel.org>]

On Wed, 24 Sep 2025 11:40:59 +0800
Julian Sun <sunjunchao@bytedance.com> wrote:

> In the kernel, long waits can trigger hung task warnings. However, some
> warnings are undesirable and unnecessary - for example, a hung task
> warning triggered when a background kworker waits for writeback
> completion during resource cleanup(like the context of
> mem_cgroup_css_free()). This kworker does not affect any user behavior
> and there is no erroneous behavior at the kernel code level, yet it
> triggers an annoying hung task warning.
> 
> To eliminate such warnings, this patch introduces
> touch_hung_task_detector() to allow some tasks ignored by hung task
> detector.
> 

Looks good to me.

Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>

Thanks,

> Signed-off-by: Julian Sun <sunjunchao@bytedance.com>
> Suggested-by: Andrew Morton <akpm@linux-foundation.org>
> Suggested-by: Lance Yang <lance.yang@linux.dev>
> ---
>  include/linux/nmi.h |  2 ++
>  kernel/hung_task.c  | 13 +++++++++++++
>  2 files changed, 15 insertions(+)
> 
> diff --git a/include/linux/nmi.h b/include/linux/nmi.h
> index cf3c6ab408aa..61fc2ad234de 100644
> --- a/include/linux/nmi.h
> +++ b/include/linux/nmi.h
> @@ -59,8 +59,10 @@ static inline void touch_all_softlockup_watchdogs(void) { }
>  
>  #ifdef CONFIG_DETECT_HUNG_TASK
>  void reset_hung_task_detector(void);
> +void touch_hung_task_detector(struct task_struct *t);
>  #else
>  static inline void reset_hung_task_detector(void) { }
> +static inline void touch_hung_task_detector(struct task_struct *t) { }
>  #endif
>  
>  /*
> diff --git a/kernel/hung_task.c b/kernel/hung_task.c
> index 8708a1205f82..6409d3d4bd36 100644
> --- a/kernel/hung_task.c
> +++ b/kernel/hung_task.c
> @@ -184,6 +184,11 @@ static inline void debug_show_blocker(struct task_struct *task)
>  }
>  #endif
>  
> +void touch_hung_task_detector(struct task_struct *t)
> +{
> +	t->last_switch_count = ULONG_MAX;
> +}
> +
>  static void check_hung_task(struct task_struct *t, unsigned long timeout)
>  {
>  	unsigned long switch_count = t->nvcsw + t->nivcsw;
> @@ -203,6 +208,10 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
>  	if (unlikely(!switch_count))
>  		return;
>  
> +	/* The task doesn't want to trigger the hung task warning. */
> +	if (unlikely(t->last_switch_count == ULONG_MAX))
> +		return;
> +
>  	if (switch_count != t->last_switch_count) {
>  		t->last_switch_count = switch_count;
>  		t->last_switch_time = jiffies;
> @@ -317,6 +326,10 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout)
>  		    !(state & TASK_WAKEKILL) &&
>  		    !(state & TASK_NOLOAD))
>  			check_hung_task(t, timeout);
> +		else if (unlikely(t->last_switch_count == ULONG_MAX)) {
> +			t->last_switch_count = t->nvcsw + t->nivcsw;
> +			t->last_switch_time = jiffies;
> +		}
>  	}
>   unlock:
>  	rcu_read_unlock();
> -- 
> 2.39.5
> 


-- 
Masami Hiramatsu (Google) <mhiramat@kernel.org>