Re: [PATCH 4/4] tracing/workqueue: Add max execution time mesurement for per worklet

[Oleg Nesterov <oleg@redhat.com>]

I have no idea how the code actually looks with these patches applied,
so please don't take my words seriously, but

On 04/28, Zhaolei wrote:
>
> @@ -24,6 +24,17 @@ struct workfunc_stats {
>  	/* Protected by cpu workqueue lock */
>  	unsigned int			inserted;
>  	unsigned int			executed;
> +
> +	/*
> +	 * save latest work_struct's pointer to use as identifier in
> +	 * probe_worklet_complete, because we can't use work_struct->...
> +	 * after worklet got executed
> +	 */
> +	void				*work;

Do we really need it ?

> @@ -143,6 +154,8 @@ found_wq:
>  	list_for_each_entry(wfnode, &node->workfunclist, list)
>  		if (wfnode->func == work->func) {
>  			wfnode->executed++;
> +			wfnode->start_time = trace_clock_global();
> +			wfnode->work = work;
>  			goto found_wf;
>  		}
>  	pr_debug("trace_workqueue: worklet not found\n");
> @@ -153,6 +166,43 @@ end:
>  	spin_unlock_irqrestore(&workqueue_cpu_stat(cpu)->lock, flags);
>  }
>
> +/* Complete of a work */
> +static void
> +probe_worklet_complete(struct task_struct *wq_thread, void *work)
> +{
> +	int cpu = cpumask_first(&wq_thread->cpus_allowed);
> +	struct cpu_workqueue_stats *node;
> +	struct workfunc_stats *wfnode;
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&workqueue_cpu_stat(cpu)->lock, flags);
> +
> +	list_for_each_entry(node, &workqueue_cpu_stat(cpu)->list, list)
> +		if (node->task == wq_thread)
> +			goto found_wq;
> +	pr_debug("trace_workqueue: workqueue not found\n");
> +	goto end;
> +
> +found_wq:
> +	list_for_each_entry(wfnode, &node->workfunclist, list) {
> +		u64 executed_time;
> +
> +		if (wfnode->work != work)
> +			continue;

Perhaps we can add node->last_work (or whatever) instead? It should be
recorded by the "entry" handler. In this case probe_worklet_complete()
doesn't need to search for this work (and it doesn't need the argument).
We know that wfnode == node->last_work.

Oleg.