RE: [PATCH] rtc: snvs: make sure clock is enabled for interrupt handle

[Anson Huang <anson.huang@nxp.com>]

Hi, Alexandre

Best Regards!
Anson Huang

> -----Original Message-----
> From: Alexandre Belloni [mailto:alexandre.belloni@bootlin.com]
> Sent: 2019年1月11日 22:32
> To: Anson Huang <anson.huang@nxp.com>
> Cc: a.zummo@towertech.it; linux-rtc@vger.kernel.org;
> linux-kernel@vger.kernel.org; dl-linux-imx <linux-imx@nxp.com>
> Subject: Re: [PATCH] rtc: snvs: make sure clock is enabled for interrupt handle
> 
> Hi,
> 
> On 11/01/2019 07:09:02+0000, Anson Huang wrote:
> > During system suspend, the SNVS RTC's clock will be disabled in noirq
> > suspend phase, but SNVS RTC's alarm interrupt could still arrive,
> > system will hang if SNVS RTC driver tries to access register without
> > clock enabled, this patch fixes the issue of this scenario.
> >
> 
> Are you sure this is the real issue? I don't think the handler can be called
> before the resume_noirq callback. Isn't the issue that your clock driver has not
> yet resumed by the time you call the rtc driver resume_noirq callback ?

Yes, since the RTC alarm is wakeup source, it is NOT being disabled during suspend
flow at all, after rtc drvier's noirq suspend callback is being called, rtc clock is disabled
in this callback, rtc alarm could arrive at this point then ARM IRQ is NOT disabled yet,
so rtc irq handler will be called, so the handler will access RTC register with RTC clock disabled
and cause system hang. The point is RTC clock is disabled in noirq suspend callback.


> 
> > Signed-off-by: Anson Huang <Anson.Huang@nxp.com>
> > ---
> >  drivers/rtc/rtc-snvs.c | 6 ++++++
> >  1 file changed, 6 insertions(+)
> >
> > diff --git a/drivers/rtc/rtc-snvs.c b/drivers/rtc/rtc-snvs.c index
> > b2483a7..0b9eff1 100644
> > --- a/drivers/rtc/rtc-snvs.c
> > +++ b/drivers/rtc/rtc-snvs.c
> > @@ -239,6 +239,9 @@ static irqreturn_t snvs_rtc_irq_handler(int irq, void
> *dev_id)
> >  	u32 lpsr;
> >  	u32 events = 0;
> >
> > +	if (data->clk)
> > +		clk_enable(data->clk);
> > +
> 
> Anyway, won't that need a clk_prepare_enable because it has been
> clk_disable_unprepare in suspend_noirq? And this is something you can not
> do in atomic context

Yes, but the prepare/unprepared can NOT be called in atomic context, since the SNVS RTC
clock on i.MX SoC is always from a clock source without prepare/unprepared needed, such as
from ipg or OSC etc., so just doing enable/disable is good enough for it.

Anson.

> 
> --
> Alexandre Belloni, Bootlin
> Embedded Linux and Kernel engineering
> https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbootl
> in.com&amp;data=02%7C01%7Canson.huang%40nxp.com%7Ca90608144e90
> 494a6f3608d677d1818a%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C
> 0%7C636828139042107931&amp;sdata=oX%2Bqnj4X3nS%2Ba88otedt9WyIy
> VhBNVl5eixVWeV44TA%3D&amp;reserved=0