| Message ID | 1399458490-11870-1-git-send-email-boris.brezillon@free-electrons.com |
|---|---|
| State | Accepted |
| Headers | show |
Boris, > -----Original Message----- > From: Boris BREZILLON [mailto:boris.brezillon@free-electrons.com] > Sent: Wednesday, May 07, 2014 6:28 AM > To: Bryan Evenson > Cc: Andrew Victor; Nicolas Ferre; Jean-Christophe Plagniol-Villard; linux-arm- > kernel@lists.infradead.org; Alessandro Zummo; rtc- > linux@googlegroups.com; linux-kernel@vger.kernel.org; Boris BREZILLON; > stable@vger.kernel.org > Subject: [PATCH v2] rtc: rtc-at91rm9200: fix uninterruptible wait for ACKUPD > > The rtc user must wait at least 1 sec between each time/calandar update > (see atmel's datasheet chapter "Updating Time/Calendar"). > > Use the SECEV interrupt (as suggested by the datasheet) to wait for the RTC > to be ready for new updates. > > Cc: <stable@vger.kernel.org> # v3.10+ > Signed-off-by: Boris BREZILLON <boris.brezillon@free-electrons.com> > Reported-by: Bryan Evenson <bevenson@melinkcorp.com> > --- > Hello, > > This patch fixes a deadlock in an uninterruptible wait when the RTC is > updated more than once every second. > AFAICT the bug is here from the beginning, but I think we should at least > backport this fix to 3.10 and the following longterm and stable releases. > > Best Regards, > > Boris Hang on, I found a problem. My system works fine on a soft reboot, but during a power cycle my system hangs. Last thing I see to print is "Uncompressing Linux... done, booting the kernel." If I removde the backup battery for the RTC my system boots just fine. Luckily this is very repeatable on my system. This looks similar to the early rtt-interrupt bug which was fixed about 7 months ago (linux-stable commits 6ce3ee9af1ad697c71766067ffa1a35352dfdcfd and 54f830b7337ed016b6708cf5a2ac297d4cee227d). I confirmed that the kernel I'm using has the relevant fix applied from 7 months ago. Did this fix reintroduce the early rtt-interrupt bug? Let me know if you would like me to test anything or more information. Regards, Bryan > > Changes since v1: > - disable SECEV interrupt when the RTC is ready to accept new time updates > > drivers/rtc/rtc-at91rm9200.c | 16 ++++++++++++++-- > 1 file changed, 14 insertions(+), 2 deletions(-) > > diff --git a/drivers/rtc/rtc-at91rm9200.c b/drivers/rtc/rtc-at91rm9200.c index > 3281c90..44fe83e 100644 > --- a/drivers/rtc/rtc-at91rm9200.c > +++ b/drivers/rtc/rtc-at91rm9200.c > @@ -48,6 +48,7 @@ struct at91_rtc_config { > > static const struct at91_rtc_config *at91_rtc_config; static > DECLARE_COMPLETION(at91_rtc_updated); > +static DECLARE_COMPLETION(at91_rtc_upd_rdy); > static unsigned int at91_alarm_year = AT91_RTC_EPOCH; static void > __iomem *at91_rtc_regs; static int irq; @@ -161,6 +162,8 @@ static int > at91_rtc_settime(struct device *dev, struct rtc_time *tm) > 1900 + tm->tm_year, tm->tm_mon, tm->tm_mday, > tm->tm_hour, tm->tm_min, tm->tm_sec); > > + wait_for_completion(&at91_rtc_upd_rdy); > + > /* Stop Time/Calendar from counting */ > cr = at91_rtc_read(AT91_RTC_CR); > at91_rtc_write(AT91_RTC_CR, cr | AT91_RTC_UPDCAL | > AT91_RTC_UPDTIM); @@ -183,7 +186,9 @@ static int > at91_rtc_settime(struct device *dev, struct rtc_time *tm) > > /* Restart Time/Calendar */ > cr = at91_rtc_read(AT91_RTC_CR); > + at91_rtc_write(AT91_RTC_SCCR, AT91_RTC_SECEV); > at91_rtc_write(AT91_RTC_CR, cr & ~(AT91_RTC_UPDCAL | > AT91_RTC_UPDTIM)); > + at91_rtc_write_ier(AT91_RTC_SECEV); > > return 0; > } > @@ -290,8 +295,10 @@ static irqreturn_t at91_rtc_interrupt(int irq, void > *dev_id) > if (rtsr) { /* this interrupt is shared! Is it ours? */ > if (rtsr & AT91_RTC_ALARM) > events |= (RTC_AF | RTC_IRQF); > - if (rtsr & AT91_RTC_SECEV) > - events |= (RTC_UF | RTC_IRQF); > + if (rtsr & AT91_RTC_SECEV) { > + complete(&at91_rtc_upd_rdy); > + at91_rtc_write_idr(AT91_RTC_SECEV); > + } > if (rtsr & AT91_RTC_ACKUPD) > complete(&at91_rtc_updated); > > @@ -413,6 +420,11 @@ static int __init at91_rtc_probe(struct > platform_device *pdev) > return PTR_ERR(rtc); > platform_set_drvdata(pdev, rtc); > > + /* enable SECEV interrupt in order to initialize at91_rtc_upd_rdy > + * completion. > + */ > + at91_rtc_write_ier(AT91_RTC_SECEV); > + > dev_info(&pdev->dev, "AT91 Real Time Clock driver.\n"); > return 0; > } > -- > 1.8.3.2
diff --git a/drivers/rtc/rtc-at91rm9200.c b/drivers/rtc/rtc-at91rm9200.c index 3281c90..44fe83e 100644 --- a/drivers/rtc/rtc-at91rm9200.c +++ b/drivers/rtc/rtc-at91rm9200.c @@ -48,6 +48,7 @@ struct at91_rtc_config { static const struct at91_rtc_config *at91_rtc_config; static DECLARE_COMPLETION(at91_rtc_updated); +static DECLARE_COMPLETION(at91_rtc_upd_rdy); static unsigned int at91_alarm_year = AT91_RTC_EPOCH; static void __iomem *at91_rtc_regs; static int irq; @@ -161,6 +162,8 @@ static int at91_rtc_settime(struct device *dev, struct rtc_time *tm) 1900 + tm->tm_year, tm->tm_mon, tm->tm_mday, tm->tm_hour, tm->tm_min, tm->tm_sec); + wait_for_completion(&at91_rtc_upd_rdy); + /* Stop Time/Calendar from counting */ cr = at91_rtc_read(AT91_RTC_CR); at91_rtc_write(AT91_RTC_CR, cr | AT91_RTC_UPDCAL | AT91_RTC_UPDTIM); @@ -183,7 +186,9 @@ static int at91_rtc_settime(struct device *dev, struct rtc_time *tm) /* Restart Time/Calendar */ cr = at91_rtc_read(AT91_RTC_CR); + at91_rtc_write(AT91_RTC_SCCR, AT91_RTC_SECEV); at91_rtc_write(AT91_RTC_CR, cr & ~(AT91_RTC_UPDCAL | AT91_RTC_UPDTIM)); + at91_rtc_write_ier(AT91_RTC_SECEV); return 0; } @@ -290,8 +295,10 @@ static irqreturn_t at91_rtc_interrupt(int irq, void *dev_id) if (rtsr) { /* this interrupt is shared! Is it ours? */ if (rtsr & AT91_RTC_ALARM) events |= (RTC_AF | RTC_IRQF); - if (rtsr & AT91_RTC_SECEV) - events |= (RTC_UF | RTC_IRQF); + if (rtsr & AT91_RTC_SECEV) { + complete(&at91_rtc_upd_rdy); + at91_rtc_write_idr(AT91_RTC_SECEV); + } if (rtsr & AT91_RTC_ACKUPD) complete(&at91_rtc_updated); @@ -413,6 +420,11 @@ static int __init at91_rtc_probe(struct platform_device *pdev) return PTR_ERR(rtc); platform_set_drvdata(pdev, rtc); + /* enable SECEV interrupt in order to initialize at91_rtc_upd_rdy + * completion. + */ + at91_rtc_write_ier(AT91_RTC_SECEV); + dev_info(&pdev->dev, "AT91 Real Time Clock driver.\n"); return 0; }
The rtc user must wait at least 1 sec between each time/calandar update (see atmel's datasheet chapter "Updating Time/Calendar"). Use the SECEV interrupt (as suggested by the datasheet) to wait for the RTC to be ready for new updates. Cc: <stable@vger.kernel.org> # v3.10+ Signed-off-by: Boris BREZILLON <boris.brezillon@free-electrons.com> Reported-by: Bryan Evenson <bevenson@melinkcorp.com> --- Hello, This patch fixes a deadlock in an uninterruptible wait when the RTC is updated more than once every second. AFAICT the bug is here from the beginning, but I think we should at least backport this fix to 3.10 and the following longterm and stable releases. Best Regards, Boris Changes since v1: - disable SECEV interrupt when the RTC is ready to accept new time updates drivers/rtc/rtc-at91rm9200.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-)