Re: [PATCH RFC 5/6] watchdog: qcom-wdt: add support to read the restart reason from IMEM

[Kathiravan Thirumoorthy]

On 4/9/2025 12:33 PM, Krzysztof Kozlowski wrote:
> > +static int  qcom_wdt_get_restart_reason(struct qcom_wdt *wdt)
> > +{
> > +	struct device_node *np;
> > +	struct resource imem;
> > +	void __iomem *base;
> > +	int ret;
> > +
> > +	np = of_find_compatible_node(NULL, NULL, "qcom,restart-reason-info");
> > +	if (!np)
> That's not how you express dependencies between devices.


As I mentioned in the bindings patch, I leveraged this from the 
qcom_pil_info.c[1]. I shall use the syscon_regmap_lookup_by_compatible() 
function.


> > +		return -ENOENT;
> > +
> > +	ret = of_address_to_resource(np, 0, &imem);
> > +	of_node_put(np);
> > +	if (ret < 0) {
> > +		dev_err(wdt->wdd.parent, "can't translate OF node address\n");
> > +		return ret;
> > +	}
> > +
> > +	base = ioremap(imem.start, resource_size(&imem));
> > +	if (!base) {
> > +		dev_err(wdt->wdd.parent, "failed to map restart reason info region\n");
> > +		return -ENOMEM;
> > +	}
> > +
> > +	memcpy_fromio(&ret, base, sizeof(ret));
> > +	iounmap(base);
> All this is wrong usage of syscon API, missing devlinks, messing up with
> other device's address space.

I shall use regmap_read() instead of memcpy_fromio().