Re: [PATCH 19/25] hw/timer/cmsdk-apb-dualtimer: Convert to use Clock input

[Peter Maydell <peter.maydell@linaro.org>]

On Fri, 22 Jan 2021 at 20:48, Luc Michel <luc@lmichel.fr> wrote:
>
> On 19:06 Thu 21 Jan     , Peter Maydell wrote:
> > Switch the CMSDK APB dualtimer device over to using its Clock input;
> > the pclk-frq property is now ignored.
> >
> > Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> > ---
> >  hw/timer/cmsdk-apb-dualtimer.c | 42 ++++++++++++++++++++++++++++++----
> >  1 file changed, 37 insertions(+), 5 deletions(-)
> >
> > diff --git a/hw/timer/cmsdk-apb-dualtimer.c b/hw/timer/cmsdk-apb-dualtimer.c
> > index 781b496037b..828127b366f 100644
> > --- a/hw/timer/cmsdk-apb-dualtimer.c
> > +++ b/hw/timer/cmsdk-apb-dualtimer.c
> > @@ -106,6 +106,22 @@ static void cmsdk_apb_dualtimer_update(CMSDKAPBDualTimer *s)
> >      qemu_set_irq(s->timerintc, timintc);
> >  }
> >
> > +static int cmsdk_dualtimermod_divisor(CMSDKAPBDualTimerModule *m)
> > +{
> > +    /* Return the divisor set by the current CONTROL.PRESCALE value */
> > +    switch (FIELD_EX32(m->control, CONTROL, PRESCALE)) {
> > +    case 0:
> > +        return 1;
> > +    case 1:
> > +        return 16;
> > +    case 2:
> > +    case 3: /* UNDEFINED, we treat like 2 (and complained when it was set) */
> > +        return 256;
> > +    default:
> > +        g_assert_not_reached();
> > +    }
> > +}
> > +
> >  static void cmsdk_dualtimermod_write_control(CMSDKAPBDualTimerModule *m,
> >                                               uint32_t newctrl)
> >  {
> > @@ -146,7 +162,7 @@ static void cmsdk_dualtimermod_write_control(CMSDKAPBDualTimerModule *m,
> >          default:
> >              g_assert_not_reached();
> >          }
> > -        ptimer_set_freq(m->timer, m->parent->pclk_frq / divisor);
> > +        ptimer_set_period_from_clock(m->timer, m->parent->timclk, divisor);
>
> Just a small cosmetic note, maybe you can use your new
> cmsdk_dualtimermod_divisor function to factor out the switch above?
> Something like:
>
> if (changed & R_CONTROL_PRESCALE_MASK) {
>     if (FIELD_EX32(newctrl, CONTROL, PRESCALE) == 3) {
>         qemu_log_mask(LOG_GUEST_ERROR,
>                       "CMSDK APB dual-timer: CONTROL.PRESCALE==0b11"
>                       " is undefined behaviour\n");
>     }
>
>     ptimer_set_period_from_clock(m->timer, m->parent->timclk,
>                                  cmsdk_dualtimermod_divisor(m));
> }

Nope, because cmsdk_dualtimermod_divisor() uses the current
m->control value, and at this point in the code we need the
divisor from the new control value which isn't in m->control yet.
I liked the slight duplication better than either having to
pass m->control in in all the other callsites or trying to
refactor the control write handling so that m->control is
updated before this point in the code.

thanks
-- PMM