Re: [PATCH 3/5] rtc: jz4740: Reset scratchpad register on power loss

[Paul Cercueil <paul@crapouillou.net>]

Hi Alexandre,

Le mar., avril 19 2022 at 21:35:35 +0200, Alexandre Belloni 
<alexandre.belloni@bootlin.com> a écrit :
> On 18/04/2022 19:49:31+0100, Paul Cercueil wrote:
>>  On power loss, reading the RTC value would fail as the scratchpad 
>> lost
>>  its magic value, until the hardware clock was set once again.
>> 
>>  To avoid that, reset the RTC value to Epoch in the probe if we 
>> detect
>>  that the scratchpad lost its magic value.
>> 
>>  Signed-off-by: Paul Cercueil <paul@crapouillou.net>
>>  ---
>>   drivers/rtc/rtc-jz4740.c | 24 +++++++++++++++++++++---
>>   1 file changed, 21 insertions(+), 3 deletions(-)
>> 
>>  diff --git a/drivers/rtc/rtc-jz4740.c b/drivers/rtc/rtc-jz4740.c
>>  index 119baf168b32..aac5f68bf626 100644
>>  --- a/drivers/rtc/rtc-jz4740.c
>>  +++ b/drivers/rtc/rtc-jz4740.c
>>  @@ -42,6 +42,9 @@
>>   /* Magic value to enable writes on jz4780 */
>>   #define JZ_RTC_WENR_MAGIC	0xA55A
>> 
>>  +/* Value written to the scratchpad to detect power losses */
>>  +#define JZ_RTC_SCRATCHPAD_MAGIC	0x12345678
>>  +
>>   #define JZ_RTC_WAKEUP_FILTER_MASK	0x0000FFE0
>>   #define JZ_RTC_RESET_COUNTER_MASK	0x00000FE0
>> 
>>  @@ -134,10 +137,11 @@ static int jz4740_rtc_ctrl_set_bits(struct 
>> jz4740_rtc *rtc, uint32_t mask,
>>   static int jz4740_rtc_read_time(struct device *dev, struct 
>> rtc_time *time)
>>   {
>>   	struct jz4740_rtc *rtc = dev_get_drvdata(dev);
>>  -	uint32_t secs, secs2;
>>  +	uint32_t secs, secs2, magic;
>>   	int timeout = 5;
>> 
>>  -	if (jz4740_rtc_reg_read(rtc, JZ_REG_RTC_SCRATCHPAD) != 0x12345678)
>>  +	magic = jz4740_rtc_reg_read(rtc, JZ_REG_RTC_SCRATCHPAD);
>>  +	if (magic != JZ_RTC_SCRATCHPAD_MAGIC)
>>   		return -EINVAL;
>> 
>>   	/* If the seconds register is read while it is updated, it can 
>> contain a
>>  @@ -169,7 +173,8 @@ static int jz4740_rtc_set_time(struct device 
>> *dev, struct rtc_time *time)
>>   	if (ret)
>>   		return ret;
>> 
>>  -	return jz4740_rtc_reg_write(rtc, JZ_REG_RTC_SCRATCHPAD, 
>> 0x12345678);
>>  +	return jz4740_rtc_reg_write(rtc, JZ_REG_RTC_SCRATCHPAD,
>>  +				    JZ_RTC_SCRATCHPAD_MAGIC);
>>   }
>> 
>>   static int jz4740_rtc_read_alarm(struct device *dev, struct 
>> rtc_wkalrm *alrm)
>>  @@ -307,6 +312,7 @@ static int jz4740_rtc_probe(struct 
>> platform_device *pdev)
>>   	struct jz4740_rtc *rtc;
>>   	unsigned long rate;
>>   	struct clk *clk;
>>  +	uint32_t magic;
>>   	int ret, irq;
>> 
>>   	rtc = devm_kzalloc(dev, sizeof(*rtc), GFP_KERNEL);
>>  @@ -369,6 +375,18 @@ static int jz4740_rtc_probe(struct 
>> platform_device *pdev)
>>   	/* Each 1 Hz pulse should happen after (rate) ticks */
>>   	jz4740_rtc_reg_write(rtc, JZ_REG_RTC_REGULATOR, rate - 1);
>> 
>>  +	magic = jz4740_rtc_reg_read(rtc, JZ_REG_RTC_SCRATCHPAD);
>>  +	if (magic != JZ_RTC_SCRATCHPAD_MAGIC) {
>>  +		/*
>>  +		 * If the scratchpad doesn't hold our magic value, then a
>>  +		 * power loss occurred. Reset to Epoch.
>>  +		 */
>>  +		struct rtc_time time;
>>  +
>>  +		rtc_time64_to_tm(0, &time);
>>  +		jz4740_rtc_set_time(dev, &time);
> 
> Don't do that, this defeats the purpose of detecting when the power is
> lost. Returning a known bogus time is the worst thing you can do here.

So what is the best thing to do then?

Cheers,
-Paul

>>  +	}
>>  +
>>   	ret = devm_rtc_register_device(rtc->rtc);
>>   	if (ret)
>>   		return ret;
>>  --
>>  2.35.1
>> 
> 
> --
> Alexandre Belloni, co-owner and COO, Bootlin
> Embedded Linux and Kernel engineering
> https://bootlin.com