Re: [PATCH 1/3] workqueue: Add workqueue_cpu_intensive tracepoint

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] 64-bit division by a constant on 32-bit architectures causes a linker error (`undefined reference to '__udivdi3'`).
--

commit 87b8b280347790d2ca859c486100f7c5e7ab2a4c
Author: Aaron Tomlin <atomlin@atomlin.com>

workqueue: Add workqueue_cpu_intensive tracepoint

This commit adds a tracepoint to record when a concurrency-managed per-CPU
work item runs continuously without sleeping for longer than the CPU
intensive threshold. It enables real-time detection and attribution of
CPU-hogging work items.

> diff --git a/kernel/workqueue.c b/kernel/workqueue.c
> index bfeef512f6dd8..957a1190aa28b 100644
> --- a/kernel/workqueue.c
> +++ b/kernel/workqueue.c
[ ... ]
> @@ -1557,9 +1558,10 @@ void wq_worker_tick(struct task_struct *task)
>  	 * double decrements. The task is releasing the CPU anyway. Let's skip.
>  	 * We probably want to make this prettier in the future.
>  	 */
> +	dur = (READ_ONCE(worker->task->se.sum_exec_runtime) - worker->current_at) /
> +	      NSEC_PER_USEC;

[Severity: Low]
Will this cause a build failure on 32-bit architectures? Using the standard
division operator to divide a 64-bit integer by a non-power-of-2 constant
(NSEC_PER_USEC) typically results in a linker error for the missing __udivdi3
compiler built-in on 32-bit targets.

Could this use div_u64() or a similar helper from linux/math64.h to prevent
the linkage error?

>  	if ((worker->flags & WORKER_NOT_RUNNING) || READ_ONCE(worker->sleeping) ||
> -	    READ_ONCE(worker->task->se.sum_exec_runtime) - worker->current_at <
> -	    wq_cpu_intensive_thresh_us * NSEC_PER_USEC)
> +	    dur < wq_cpu_intensive_thresh_us)
>  		return;
>  
>  	raw_spin_lock(&pool->lock);

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260829230517.42468-1-atomlin@atomlin.com?part=1