Re: [PATCH 01/15] tick/nohz: Remove duplicate between tick_nohz_switch_to_nohz() and tick_setup_sched_timer()

[Frederic Weisbecker]

On Thu, Jan 25, 2024 at 10:12:20AM +0100, Thomas Gleixner wrote:
> On Wed, Jan 24 2024 at 18:04, Frederic Weisbecker wrote:
> >  
> >  /**
> >   * tick_setup_sched_timer - setup the tick emulation timer
> > + * @mode: tick_nohz_mode to setup for
> >   */
> > -void tick_setup_sched_timer(void)
> > +void tick_setup_sched_timer(int mode)
> >  {
> >  	struct tick_sched *ts = this_cpu_ptr(&tick_cpu_sched);
> > -	ktime_t now = ktime_get();
> >  
> >  	/* Emulate tick processing via per-CPU hrtimers: */
> >  	hrtimer_init(&ts->sched_timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS_HARD);
> > -	ts->sched_timer.function = tick_nohz_highres_handler;
> > +#ifdef CONFIG_HIGH_RES_TIMERS
> > +	if (mode == NOHZ_MODE_HIGHRES)
> > +		ts->sched_timer.function = tick_nohz_highres_handler;
> > +#endif
> 
> Adding
> 
> #define tick_nohz_highres_handler NULL
> 
> for CONFIG_HIGH_RES_TIMERS=n somewhere allows you to write that as:
> 
> 	if (IS_ENABLED(CONFIG_HIGH_RES_TIMERS) && mode == NOHZ_MODE_HIGHRES)
> 		ts->sched_timer.function = tick_nohz_highres_handler;

So the next patch does:

if (mode == NOHZ_MODE_HIGHRES)
    ts->sched_timer.function = tick_nohz_highres_handler;

Because that condition only happens in CONFIG_HIGH_RES_TIMERS anyway
and the function field exists in any case.

I can move that to the current patch though.

>     
> > +	hrtimer_forward_now(&ts->sched_timer, TICK_NSEC);
> > +	if (mode == NOHZ_MODE_HIGHRES)
> 
> Wants a (IS_ENABLED(HIGRES) && mode == ...) no?

Hmm, indeed hrtimer_start_range_ns() called by hrtimer_start_expires()
is only defined if CONFIG_HIGH_RES_TIMERS=y. I'll resend.

Thanks.

> 
> Thanks,
> 
>         tglx