RTC: RK808: fix the rtc time reading issue

After we set the GET_TIME bit, the rtc time couldn't be read immediately,
we should wait up to 31.25 us, about one cycle of 32khz. Otherwise reading
RTC time will return a old time. If clear the GET_TIME bit after setting,
the time of i2c transfer certainly more than 31.25us.

Signed-off-by: Chris Zhong <zyw@rock-chips.com>

---

 drivers/rtc/rtc-rk808.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

On Tue, Jan 13, 2015 at 6:43 PM, Chris Zhong <zyw@rock-chips.com> wrote:
> After we set the GET_TIME bit, the rtc time couldn't be read immediately,
> we should wait up to 31.25 us, about one cycle of 32khz. Otherwise reading
> RTC time will return a old time. If clear the GET_TIME bit after setting,
> the time of i2c transfer certainly more than 31.25us.
>
> Signed-off-by: Chris Zhong <zyw@rock-chips.com>
>
> ---
>
>  drivers/rtc/rtc-rk808.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/rtc/rtc-rk808.c b/drivers/rtc/rtc-rk808.c
> index df42257..8dae322 100644
> --- a/drivers/rtc/rtc-rk808.c
> +++ b/drivers/rtc/rtc-rk808.c
> @@ -67,15 +67,20 @@ static int rk808_rtc_readtime(struct device *dev, struct rtc_time *tm)
>         /* Force an update of the shadowed registers right now */
>         ret = regmap_update_bits(rk808->regmap, RK808_RTC_CTRL_REG,
>                                  BIT_RTC_CTRL_REG_RTC_GET_TIME,
> -                                0);
> +                                BIT_RTC_CTRL_REG_RTC_GET_TIME);
>         if (ret) {
>                 dev_err(dev, "Failed to update bits rtc_ctrl: %d\n", ret);
>                 return ret;
>         }
>
> +       /* After we set the GET_TIME bit, the rtc time couldn't be read
> +        * immediately, we should wait up to 31.25 us, about one cycle of
> +        * 32khz. If we clear the GET_TIME bit here, the time of i2c transfer
> +        * certainly more than 31.25us.
> +        */

Chris, it looks like you swapped the set and the clear of this bit,
and you're relying on the fact that the i2c transaction takes a
certain amount of time after the RTC_GET_TIME BIT is set.   I'm not
sure how long it actually takes, but why not just put in a usleep()
for the minimum wait time?

>         ret = regmap_update_bits(rk808->regmap, RK808_RTC_CTRL_REG,
>                                  BIT_RTC_CTRL_REG_RTC_GET_TIME,
> -                                BIT_RTC_CTRL_REG_RTC_GET_TIME);
> +                                0);
>         if (ret) {
>                 dev_err(dev, "Failed to update bits rtc_ctrl: %d\n", ret);
>                 return ret;
> --
> 1.9.1
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Sonny,

> Chris, it looks like you swapped the set and the clear of this bit,
> and you're relying on the fact that the i2c transaction takes a
> certain amount of time after the RTC_GET_TIME BIT is set.   I'm not
> sure how long it actually takes, but why not just put in a usleep()
> for the minimum wait time?

I think we are safe.

At 400kHz (the max speed of this part) each bit can be transferred no
faster than 2.5us.  In order to do a valid i2c transaction we need to
_at least_ write the address of the device and the data onto the bus,
which is 16 bits.  16 * 2.5us = 40us.  That's above the 31.25us

Personally I think what Chris has is fine, with the comment.

-Doug

Chris,

On Tue, Jan 13, 2015 at 6:43 PM, Chris Zhong <zyw@rock-chips.com> wrote:
> +       /* After we set the GET_TIME bit, the rtc time couldn't be read
> +        * immediately, we should wait up to 31.25 us, about one cycle of
> +        * 32khz. If we clear the GET_TIME bit here, the time of i2c transfer
> +        * certainly more than 31.25us.
> +        */

No idea what Andrew thinks of this, but other multiline comments in
this file look slightly different (there's a line containing just /*).


>         ret = regmap_update_bits(rk808->regmap, RK808_RTC_CTRL_REG,
>                                  BIT_RTC_CTRL_REG_RTC_GET_TIME,
> -                                BIT_RTC_CTRL_REG_RTC_GET_TIME);
> +                                0);
>         if (ret) {
>                 dev_err(dev, "Failed to update bits rtc_ctrl: %d\n", ret);
>                 return ret;

Above is just a nit.  I'll leave it up to Andrew to say whether he
wants you to spin, or he'll fix it up, or he doesn't care and he'll
land it as-is.

Reviewed-by: Doug Anderson <dianders@chromium.org>

On Wed, Jan 14, 2015 at 10:36 AM, Doug Anderson <dianders@chromium.org> wrote:
> Sonny,
>
>> Chris, it looks like you swapped the set and the clear of this bit,
>> and you're relying on the fact that the i2c transaction takes a
>> certain amount of time after the RTC_GET_TIME BIT is set.   I'm not
>> sure how long it actually takes, but why not just put in a usleep()
>> for the minimum wait time?
>
> I think we are safe.
>
> At 400kHz (the max speed of this part) each bit can be transferred no
> faster than 2.5us.  In order to do a valid i2c transaction we need to
> _at least_ write the address of the device and the data onto the bus,
> which is 16 bits.  16 * 2.5us = 40us.  That's above the 31.25us
>
> Personally I think what Chris has is fine, with the comment.

Ok, I'm fine with that if we're sure it's slow enough.  Comment
explaining would certainly help.

>
> -Doug