RE: [PATCH 3/7] watchdog: imx7ulp_wdt: Check CMD32EN in wdog init

[Alice Guo (OSS)]

> -----Original Message-----
> From: Guenter Roeck <groeck7@xxxxxxxxx> On Behalf Of Guenter Roeck
> Sent: Monday, August 22, 2022 10:06 PM
> To: Alice Guo (OSS) <alice.guo@xxxxxxxxxxx>
> Cc: wim@xxxxxxxxxxxxxxxxxx; shawnguo@xxxxxxxxxx;
> s.hauer@xxxxxxxxxxxxxx; festevam@xxxxxxxxx; kernel@xxxxxxxxxxxxxx;
> dl-linux-imx <linux-imx@xxxxxxx>; linux-watchdog@xxxxxxxxxxxxxxx;
> linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH 3/7] watchdog: imx7ulp_wdt: Check CMD32EN in wdog
> init
> 
> On Tue, Aug 16, 2022 at 12:36:39PM +0800, Alice Guo (OSS) wrote:
> > From: Ye Li <ye.li@xxxxxxx>
> >
> > When bootloader has enabled the CMD32EN bit, switch to use 32bits
> > unlock command to unlock the CS register. Using 32bits command will
> > help on avoiding 16 bus cycle window violation for two 16 bits
> > commands.
> >
> > Signed-off-by: Ye Li <ye.li@xxxxxxx>
> > Signed-off-by: Alice Guo <alice.guo@xxxxxxx>
> > Reviewed-by: Jacky Bai <ping.bai@xxxxxxx>
> > Acked-by: Jason Liu <jason.hui.liu@xxxxxxx>
> > ---
> >  drivers/watchdog/imx7ulp_wdt.c | 15 ++++++++++-----
> >  1 file changed, 10 insertions(+), 5 deletions(-)
> >
> > diff --git a/drivers/watchdog/imx7ulp_wdt.c
> > b/drivers/watchdog/imx7ulp_wdt.c index b8ac0cb04d2f..a0f6b8cea78f
> > 100644
> > --- a/drivers/watchdog/imx7ulp_wdt.c
> > +++ b/drivers/watchdog/imx7ulp_wdt.c
> > @@ -180,11 +180,16 @@ static int imx7ulp_wdt_init(void __iomem *base,
> > unsigned int timeout)
> >
> >  	local_irq_disable();
> >
> > -	mb();
> > -	/* unlock the wdog for reconfiguration */
> > -	writel_relaxed(UNLOCK_SEQ0, base + WDOG_CNT);
> > -	writel_relaxed(UNLOCK_SEQ1, base + WDOG_CNT);
> > -	mb();
> > +	val = readl(base + WDOG_CS);
> > +	if (val & WDOG_CS_CMD32EN) {
> > +		writel(UNLOCK, base + WDOG_CNT);
> > +	} else {
> > +		mb();
> > +		/* unlock the wdog for reconfiguration */
> > +		writel_relaxed(UNLOCK_SEQ0, base + WDOG_CNT);
> > +		writel_relaxed(UNLOCK_SEQ1, base + WDOG_CNT);
> > +		mb();
> 
> Now this is intermixing writel() with writel_relaxed(), making the code all but
> impossible to understand.
> 
> Guenter

Hi Guenter,

Intermixing writel() with writel_relaxed() is unavoidable here. Because there cannot be a memory barrier between writing UNLOCK_SEQ0 and writing UNLOCK_SEQ1. This may be determined by hardware design.

Best Regards,
Alice Guo
 
> > +	}
> >
> >  	ret = imx7ulp_wdt_wait(base, WDOG_CS_ULK);
> >  	if (ret)
> > --
> > 2.17.1
> >