Re: [PATCH v2 8/9] rtc: mt6397: fix alarm register overwrite

[mtk14576]

Hi,

On Mon, 2019-03-11 at 13:50 -0700, Sean Wang wrote:
> Hi,
> 
> On Sun, Mar 10, 2019 at 8:49 PM Hsin-Hsiung Wang
> <hsin-hsiung.wang@xxxxxxxxxxxx> wrote:
> >
> > From: Ran Bi <ran.bi@xxxxxxxxxxxx>
> >
> > Alarm registers high byte was reserved for other functions.
> > This add mask in alarm registers operation functions.
> > This also fix error condition in interrupt handler.
> >
> > Fixes: fc2979118f3f ("rtc: mediatek: Add MT6397 RTC driver")
> >
> 
> add a  Cc: stable@xxxxxxxxxxxxxxx  here to apply to all stable kernels
> for the critical fixup patch.
> 
> > Signed-off-by: Ran Bi <ran.bi@xxxxxxxxxxxx>
> > ---
> >  drivers/rtc/rtc-mt6397.c | 47 +++++++++++++++++++++++++++++++++--------------
> >  1 file changed, 33 insertions(+), 14 deletions(-)
> >
> > diff --git a/drivers/rtc/rtc-mt6397.c b/drivers/rtc/rtc-mt6397.c
> > index e9a25ec..f85f1fc 100644
> > --- a/drivers/rtc/rtc-mt6397.c
> > +++ b/drivers/rtc/rtc-mt6397.c
> > @@ -55,6 +55,14 @@
> >
> >  #define RTC_AL_SEC             0x0018
> >
> > +#define RTC_AL_SEC_MASK                0x003f
> > +#define RTC_AL_MIN_MASK                0x003f
> > +#define RTC_AL_HOU_MASK                0x001f
> > +#define RTC_AL_DOM_MASK                0x001f
> > +#define RTC_AL_DOW_MASK                0x0007
> > +#define RTC_AL_MTH_MASK                0x000f
> > +#define RTC_AL_YEA_MASK                0x007f
> > +
> >  #define RTC_PDN2               0x002e
> >  #define RTC_PDN2_PWRON_ALARM   BIT(4)
> >
> > @@ -111,7 +119,7 @@ static irqreturn_t mtk_rtc_irq_handler_thread(int irq, void *data)
> >                 irqen = irqsta & ~RTC_IRQ_EN_AL;
> >                 mutex_lock(&rtc->lock);
> >                 if (regmap_write(rtc->regmap, rtc->addr_base + RTC_IRQ_EN,
> > -                                irqen) < 0)
> > +                                irqen) == 0)
> >                         mtk_rtc_write_trigger(rtc);
> >                 mutex_unlock(&rtc->lock);
> >
> > @@ -233,12 +241,12 @@ static int mtk_rtc_read_alarm(struct device *dev, struct rtc_wkalrm *alm)
> >         alm->pending = !!(pdn2 & RTC_PDN2_PWRON_ALARM);
> >         mutex_unlock(&rtc->lock);
> >
> > -       tm->tm_sec = data[RTC_OFFSET_SEC];
> > -       tm->tm_min = data[RTC_OFFSET_MIN];
> > -       tm->tm_hour = data[RTC_OFFSET_HOUR];
> > -       tm->tm_mday = data[RTC_OFFSET_DOM];
> > -       tm->tm_mon = data[RTC_OFFSET_MTH];
> > -       tm->tm_year = data[RTC_OFFSET_YEAR];
> > +       tm->tm_sec = data[RTC_OFFSET_SEC] & RTC_AL_SEC_MASK;
> > +       tm->tm_min = data[RTC_OFFSET_MIN] & RTC_AL_MIN_MASK;
> > +       tm->tm_hour = data[RTC_OFFSET_HOUR] & RTC_AL_HOU_MASK;
> > +       tm->tm_mday = data[RTC_OFFSET_DOM] & RTC_AL_DOM_MASK;
> > +       tm->tm_mon = data[RTC_OFFSET_MTH] & RTC_AL_MTH_MASK;
> > +       tm->tm_year = data[RTC_OFFSET_YEAR] & RTC_AL_YEA_MASK;
> >
> >         tm->tm_year += RTC_MIN_YEAR_OFFSET;
> >         tm->tm_mon--;
> > @@ -259,14 +267,25 @@ static int mtk_rtc_set_alarm(struct device *dev, struct rtc_wkalrm *alm)
> >         tm->tm_year -= RTC_MIN_YEAR_OFFSET;
> >         tm->tm_mon++;
> >
> > -       data[RTC_OFFSET_SEC] = tm->tm_sec;
> > -       data[RTC_OFFSET_MIN] = tm->tm_min;
> > -       data[RTC_OFFSET_HOUR] = tm->tm_hour;
> > -       data[RTC_OFFSET_DOM] = tm->tm_mday;
> > -       data[RTC_OFFSET_MTH] = tm->tm_mon;
> > -       data[RTC_OFFSET_YEAR] = tm->tm_year;
> > -
> >         mutex_lock(&rtc->lock);
> > +       ret = regmap_bulk_read(rtc->regmap, rtc->addr_base + RTC_AL_SEC,
> > +                              data, RTC_OFFSET_COUNT);
> > +       if (ret < 0)
> > +               goto exit;
> > +
> 
> Why did we need an additional regmap_bulk_read here? I would suppose
> data[RTC_OFFSET_*] & ~(RTC_AL_*_MASK) is always equal to 0.

In bootloader, RTC init part need to operate these RTC_AL_* registers
high byte for RTC operating mode settings and data storage, while
RTC_AL_* low byte was used as alarm register here. So we need an
additional regmap_bulk_read here to keep the RTC settings correct.

> 
> It might be another fixup since change is not being mentioned in the
> git message.
> 
> > +       data[RTC_OFFSET_SEC] = ((data[RTC_OFFSET_SEC] & ~(RTC_AL_SEC_MASK)) |
> > +                               (tm->tm_sec & RTC_AL_SEC_MASK));
> > +       data[RTC_OFFSET_MIN] = ((data[RTC_OFFSET_MIN] & ~(RTC_AL_MIN_MASK)) |
> > +                               (tm->tm_min & RTC_AL_MIN_MASK));
> > +       data[RTC_OFFSET_HOUR] = ((data[RTC_OFFSET_HOUR] & ~(RTC_AL_HOU_MASK)) |
> > +                               (tm->tm_hour & RTC_AL_HOU_MASK));
> > +       data[RTC_OFFSET_DOM] = ((data[RTC_OFFSET_DOM] & ~(RTC_AL_DOM_MASK)) |
> > +                               (tm->tm_mday & RTC_AL_DOM_MASK));
> > +       data[RTC_OFFSET_MTH] = ((data[RTC_OFFSET_MTH] & ~(RTC_AL_MTH_MASK)) |
> > +                               (tm->tm_mon & RTC_AL_MTH_MASK));
> > +       data[RTC_OFFSET_YEAR] = ((data[RTC_OFFSET_YEAR] & ~(RTC_AL_YEA_MASK)) |
> > +                               (tm->tm_year & RTC_AL_YEA_MASK));
> > +
> >         if (alm->enabled) {
> >                 ret = regmap_bulk_write(rtc->regmap,
> >                                         rtc->addr_base + RTC_AL_SEC,
> > --
> > 1.9.1
> >
> >
> > _______________________________________________
> > Linux-mediatek mailing list
> > Linux-mediatek@xxxxxxxxxxxxxxxxxxx
> > http://lists.infradead.org/mailman/listinfo/linux-mediatek