On Tue, Mar 31, 2020 at 12:44 PM Stefan Roese <s...@denx.de> wrote:
>
> On 31.03.20 08:08, Rayagonda Kokatanur wrote:
> > From: Bharat Kumar Reddy Gooty <bharat.go...@broadcom.com>
> >
> > Get the watchdog default timeout value and clock from the DTS file
>
> The default timeout is already read from the DT. Please see below...
Thank you, will fix this.
>
> > Signed-off-by: Bharat Kumar Reddy Gooty <bharat.go...@broadcom.com>
> > Signed-off-by: Rayagonda Kokatanur <rayagonda.kokata...@broadcom.com>
> > ---
> > drivers/watchdog/sp805_wdt.c | 36 ++++++++++++++++++++++++++++++++++--
> > 1 file changed, 34 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/watchdog/sp805_wdt.c b/drivers/watchdog/sp805_wdt.c
> > index ca3ccbe76c..e8f54d6804 100644
> > --- a/drivers/watchdog/sp805_wdt.c
> > +++ b/drivers/watchdog/sp805_wdt.c
> > @@ -34,8 +34,31 @@ DECLARE_GLOBAL_DATA_PTR;
> >
> > struct sp805_wdt_priv {
> > void __iomem *reg;
> > + u32 timeout_msec;
> > + u32 clk_mhz;
> > };
> >
> > +static u32 msec_to_ticks(struct udevice *dev)
> > +{
> > + u32 timeout_msec;
> > + u32 msec;
> > + struct sp805_wdt_priv *priv = dev_get_priv(dev);
> > +
> > + timeout_msec = env_get_ulong("wdt_timeout_msec", 10, 0);
> > + if (timeout_msec) {
> > + dev_dbg(dev, "Overriding timeout :%u\n", timeout_msec);
> > + msec = timeout_msec;
> > + } else {
> > + msec = priv->timeout_msec;
> > + }
>
> Why is this overriding via environment needed? BTW: You should mention
> such a change in the commit text as well.
This code will be removed as default timeout is already read in include/wdt.h
>
> > +
> > + timeout_msec = (msec / 2) * (priv->clk_mhz / 1000);
> > +
> > + dev_dbg(dev, "ticks :%u\n", timeout_msec);
> > +
> > + return timeout_msec;
> > +}
> > +
> > static int sp805_wdt_reset(struct udevice *dev)
> > {
> > struct sp805_wdt_priv *priv = dev_get_priv(dev);
> > @@ -63,8 +86,11 @@ static int sp805_wdt_start(struct udevice *dev, u64
> > timeout, ulong flags)
> > * set 120s, the gd->bus_clk is less than 1145MHz, the load_value will
> > * not overflow.
> > */
> > - load_value = (gd->bus_clk) /
> > - (2 * 1000 * SYS_FSL_WDT_CLK_DIV) * load_time;
> > + if (gd->bus_clk)
> > + load_value = (gd->bus_clk) /
> > + (2 * 1000 * SYS_FSL_WDT_CLK_DIV) * load_time;
> > + else /* platform provide clk */
> > + load_value = msec_to_ticks(dev);
>
> Nitpicking: Please use paranthesis on multi-line statements.
Thank you fix this.
>
> > writel(UNLOCK, priv->reg + WDTLOCK);
> > writel(load_value, priv->reg + WDTLOAD);
> > @@ -110,6 +136,12 @@ static int sp805_wdt_ofdata_to_platdata(struct udevice
> > *dev)
> > if (IS_ERR(priv->reg))
> > return PTR_ERR(priv->reg);
> >
> > + if (dev_read_u32(dev, "timeout-msec", &priv->timeout_msec))
> > + return -ENODATA;
>
> This does not seem to be an official DT binding. At least I was unable
> to find it in the latest Linux tree.
>
> You are aware that the WDT core code already reads the WDT timeout from
> the DT using the official "timeout-sec" property? Do you need a higher
> graned timeout value (ms vs s)?
Thank you for pointing this, will update my dt file.
>
> Thanks,
> Stefan
