Hi, Alexandre, On 02.09.2024 17:49, claudiu beznea wrote: > Hi, Alexandre, > > On 31.08.2024 01:25, Alexandre Belloni wrote: >> On 30/08/2024 16:02:12+0300, Claudiu wrote: >>> + priv->rtc_dev->range_min = mktime64(2000, 1, 1, 0, 0, 0); >> >> RTC_TIMESTAMP_BEGIN_2000 > > OK > >> >>> + priv->rtc_dev->range_max = mktime64(2099, 12, 31, 23, 59, 59); >> >> RTC_TIMESTAMP_END_2099 > > OK > >> >>> + >>> + return devm_rtc_register_device(priv->rtc_dev); >>> +} >>> + >>> +static void rtca3_remove(struct platform_device *pdev) >>> +{ >>> + struct rtca3_priv *priv = platform_get_drvdata(pdev); >>> + >>> + guard(spinlock_irqsave)(&priv->lock); >>> + >>> + /* Disable alarm, periodic interrupts. */ >>> + rtca3_alarm_irq_set_helper(priv, RTCA3_RCR1_AIE | RTCA3_RCR1_PIE, 0); >> >> Why do you disable alarms on driver remove? I think you need to add a >> comment if this is because it can't system up, else this is a bad >> practice. > > The RTC cannot power on the system after a power off. It can't also resume > it from a deep sleep state (when only the SoC area where the RTC resides > remains power on (there is no way to signal from RTC to the power supply > chain that an alarm happened)). It can only wake it up from s2idle mode > where all SoC components remains powered on. FTR, this is still valid. > > Also, w/o this change the RTC remains blocked under the following scenarios > if the interrupts are not disabled in remove: This intrigued me and did some further investigation. I found that something is wrong on the driver as described bellow. The failures described in the previous emails were due to the fact that the RTC counter clock was requested by the driver as devres managed resource. This is the clock that feeds the RTC counting logic. With the bellow diff applied on top of this series scenarios 1/ and 2/ described in the previous email works just fine. For scenario 3/ the system is not powered up (as expected) but there are no more failures on the next boots. diff --git a/drivers/rtc/rtc-renesas-rtca3.c b/drivers/rtc/rtc-renesas-rtca3.c index f908d2a1017d..c9adcadc58c0 100644 --- a/drivers/rtc/rtc-renesas-rtca3.c +++ b/drivers/rtc/rtc-renesas-rtca3.c @@ -747,10 +747,14 @@ static int rtca3_probe(struct platform_device *pdev) if (ret) return ret; - clk = devm_clk_get_enabled(dev, "counter"); + clk = devm_clk_get(dev, "counter"); if (IS_ERR(clk)) return PTR_ERR(clk); + ret = clk_prepare_enable(clk); + if (ret) + return ret; + spin_lock_init(&priv->lock); atomic_set(&priv->alrm_sstep, RTCA3_ALRM_SSTEP_DONE); init_completion(&priv->set_alarm_completion); @@ -783,7 +787,7 @@ static void rtca3_remove(struct platform_device *pdev) guard(spinlock_irqsave)(&priv->lock); - rtca3_alarm_irq_set_helper(priv, RTCA3_RCR1_AIE | RTCA3_RCR1_PIE, 0); + //rtca3_alarm_irq_set_helper(priv, RTCA3_RCR1_AIE | RTCA3_RCR1_PIE, 0); } Question is: the RTC counter clock should stay on when the driver is unbinded, right? Thank you, Claudiu Beznea > > 1/ Configure wake alarm and unbind the RTC driver with the following commands: > # echo +10 > /sys/class/rtc/rtc0/wakealarm > # echo /sys/bus/platform/drivers/rtc-rtca3/1004ec00.rtc > unbind > # sleep 12 > # echo /sys/bus/platform/drivers/rtc-rtca3/1004ec00.rtc > bind > > When rebinding the re-configuration of the RTC device times out: > [ 121.854190] rtc-rtca3 1004ec00.rtc: error -ETIMEDOUT: Failed to setup > the RTC! > [ 121.861511] rtc-rtca3 1004ec00.rtc: probe with driver rtc-rtca3 failed > with error -110 > -sh: echo: write error: Connection timed out > > 2/ Configure wake alarm, unbind the RTC driver and switch to s2idle with > the following commands: > # echo s2idle > /sys/power/mem_sleep > # echo +10 > /sys/class/rtc/rtc0/wakealarm > # echo /sys/bus/platform/drivers/rtc-rtca/31004ec00.rtc > unbind > # echo mem > /sys/power/state > # #system is resumed by non RTC wakeup source (as the RTC alarm is not > working anymore in this case) > # echo /sys/bus/platform/drivers/rtc-rtca/1004ec00.rtc > bind > > The system is not waked up from RTC alarm (as expected) and the rebinding > fails again: > > [ 172.483688] rtc-rtca3 1004ec00.rtc: error -ETIMEDOUT: Failed to setup > the RTC! > [ 172.491003] rtc-rtca3 1004ec00.rtc: probe with driver rtc-rtca3 failed > with error -110 > -sh: echo: write error: Connection timed out > > 3/ configure the RTC alarm, unbind and power off (with the following commands): > # echo +60 > /sys/class/rtc/rtc0/wakealarm > # echo /sys/bus/platform/drivers/rtc-rtca/1004ec00.rtc > unbind > # poweroff > > The system is not started after 60 seconds and at the next reboot the RTC > configuration on probe is failing the same: > > [ 0.292068] rtc-rtca3 1004ec00.rtc: error -ETIMEDOUT: Failed to setup > the RTC! > [ 0.292182] rtc-rtca3 1004ec00.rtc: probe with driver rtc-rtca3 failed > with error -110 > > In all scenarios the RTC is recovered only if removing/re-applying the > power to the SoC area where it resides. > > These tests were done with the patches in this series and then I tried it > with the following diff on top of the patches in this series. The results > were the same: > > diff --git a/drivers/rtc/rtc-renesas-rtca3.c b/drivers/rtc/rtc-renesas-rtca3.c > index 822c055b6e4d..720fdac3adc6 100644 > --- a/drivers/rtc/rtc-renesas-rtca3.c > +++ b/drivers/rtc/rtc-renesas-rtca3.c > @@ -586,7 +586,7 @@ static int rtca3_initial_setup(struct clk *clk, struct > rtca3_priv *priv) > usleep_range(sleep_us, sleep_us + 10); > > /* Disable alarm and carry interrupts. */ > - mask = RTCA3_RCR1_AIE | RTCA3_RCR1_CIE; > + mask = RTCA3_RCR1_AIE | RTCA3_RCR1_CIE | RTCA3_RCR1_PIE; > ret = rtca3_alarm_irq_set_helper(priv, mask, 0); > if (ret) > return ret; > @@ -784,7 +784,7 @@ static void rtca3_remove(struct platform_device *pdev) > guard(spinlock_irqsave)(&priv->lock); > > /* Disable alarm, periodic interrupts. */ > - rtca3_alarm_irq_set_helper(priv, RTCA3_RCR1_AIE | RTCA3_RCR1_PIE, 0); > + //rtca3_alarm_irq_set_helper(priv, RTCA3_RCR1_AIE | RTCA3_RCR1_PIE, 0); > } > > Thank you, > Claudiu Beznea >