kernel/latencytop.c: remove unnecessary checks for latencytop_enabled

1. In latencytop source codes, we only have such calling chain:

account_scheduler_latency(struct task_struct *task, int usecs, int inter)
{
        if (unlikely(latencytop_enabled)) /* the outtermost check */
                __account_scheduler_latency(task, usecs, inter);
}
__account_scheduler_latency
    account_global_scheduler_latency
        if (!latencytop_enabled)

So, the inner check for latencytop_enabled is not necessary at all.

2. In clear_all_latency_tracing and now is called
   clear_tsk_latency_tracing the check for latencytop_enabled is redundant
   and buggy to some extent.

   We have no reason to refuse clearing the /proc/$pid/latency if
   latencytop_enabled is set to 0, considering that if we use latencytop
   manually by echo 0 > /proc/sys/kernel/latencytop, then we want to clear
   /proc/$pid/latency and failed.

   Also we don't have such check in brother function
   clear_global_latency_tracing.

Notes: These changes are only visible to users who set
   CONFIG_LATENCYTOP and won't change user tool latencytop's behavior.

Link: http://lkml.kernel.org/r/20190226114602.16902-2-linf@wangsu.com
Signed-off-by: Lin Feng <linf@wangsu.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Fabian Frederick <fabf@skynet.be>
Cc: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/latencytop.c

@@ -71,9 +71,6 @@ void clear_all_latency_tracing(struct task_struct *p)
 {
 	unsigned long flags;
 
-	if (!latencytop_enabled)
-		return;
-
 	raw_spin_lock_irqsave(&latency_lock, flags);
 	memset(&p->latency_record, 0, sizeof(p->latency_record));
 	p->latency_record_count = 0;
@@ -96,9 +93,6 @@ account_global_scheduler_latency(struct task_struct *tsk,
 	int firstnonnull = MAXLR + 1;
 	int i;
 
-	if (!latencytop_enabled)
-		return;
-
 	/* skip kernel threads for now */
 	if (!tsk->mm)
 		return;