| Message ID | 1446827876-31613-1-git-send-email-u.kleine-koenig@pengutronix.de |
|---|---|
| State | Accepted |
| Headers | show |
Hi, On 06/11/2015 at 17:37:56 +0100, Uwe Kleine-König wrote : > When the chip increments the YEAR register and it already holds > bin2bcd(99) it reads as 0 afterwards. With this behaviour the last valid > day (without trickery) that has a representation is 2099-12-31 23:59:59. > So refuse to write later dates. > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> > --- > Hello, > > When setting the YEAR register to bin2bcd(100) == 0xa0 it increments > just fine up to bin2bcd(159) == 0xf9 but these values are forbidden in > the manual. But this doesn't help to expand (or shift) the range of > supported dates. > > An alternative to this patch would be to save tm_year % 100 because > that's what happens when waiting long enough anyhow. Would that be > better? Do we already care at all for year-2100-problems? > There is no way to tell whether the date is fine after February 2100 so I will take the patch as is.
diff --git a/drivers/rtc/rtc-pcf8523.c b/drivers/rtc/rtc-pcf8523.c index 4cdb64be061b..261d65e67c8b 100644 --- a/drivers/rtc/rtc-pcf8523.c +++ b/drivers/rtc/rtc-pcf8523.c @@ -219,6 +219,17 @@ static int pcf8523_rtc_set_time(struct device *dev, struct rtc_time *tm) u8 regs[8]; int err; + /* + * The hardware can only store values between 0 and 99 in it's YEAR + * register (with 99 overflowing to 0 on increment). + * After 2100-02-28 we could start interpreting the year to be in the + * interval [2100, 2199], but there is no path to switch in a smooth way + * because the chip handles YEAR=0x00 (and the out-of-spec + * YEAR=0xa0) as a leap year, but 2100 isn't. + */ + if (tm->tm_year < 100 || tm->tm_year >= 200) + return -EINVAL; + err = pcf8523_stop_rtc(client); if (err < 0) return err;
When the chip increments the YEAR register and it already holds bin2bcd(99) it reads as 0 afterwards. With this behaviour the last valid day (without trickery) that has a representation is 2099-12-31 23:59:59. So refuse to write later dates. Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> --- Hello, When setting the YEAR register to bin2bcd(100) == 0xa0 it increments just fine up to bin2bcd(159) == 0xf9 but these values are forbidden in the manual. But this doesn't help to expand (or shift) the range of supported dates. An alternative to this patch would be to save tm_year % 100 because that's what happens when waiting long enough anyhow. Would that be better? Do we already care at all for year-2100-problems? Best regards Uwe drivers/rtc/rtc-pcf8523.c | 11 +++++++++++ 1 file changed, 11 insertions(+)